1<!DOCTYPE html> 2<html> 3 4 <head> 5 6 <meta charset="utf-8"> 7<meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no"> 8<meta http-equiv="X-UA-Compatible" content="IE=edge"> 9 10<title> 11 12 Harshavardhan Kamarthi 13 14 15</title> 16<meta name="description" content="Harsha's "About" page. 17"> 18 19<!-- Open Graph --> 20 21 22<!-- Bootstrap & MDB --> 23<link href="https://stackpath.bootstrapcdn.com/bootstrap/4.5.2/css/bootstrap.min.css" rel="stylesheet" integrity="sha512-MoRNloxbStBcD8z3M/2BmnT+rg4IsMxPkXaGh2zD6LGNNFE80W3onsAhRcMAMrSoyWL9xD7Ert0men7vR8LUZg==" crossorigin="anonymous"> 24<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/mdbootstrap/4.19.1/css/mdb.min.css" integrity="sha512-RO38pBRxYH3SoOprtPTD86JFOclM51/XTIdEPh5j8sj4tp8jmQIx26twG52UaLi//hQldfrh7e51WzP9wuP32Q==" crossorigin="anonymous" /> 25 26<!-- Fonts & Icons --> 27<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/5.14.0/css/all.min.css" integrity="sha512-1PKOgIY59xJ8Co8+NE6FZ+LOAZKjy+KY8iq0G4B3CyeY6wYHN3yt9PW0XpSriVlkMXe40PTKnXrLnZ9+fkDaog==" crossorigin="anonymous"> 28<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/academicons/1.9.0/css/academicons.min.css" integrity="sha512-W4yqoT1+8NLkinBLBZko+dFB2ZbHsYLDdr50VElllRcNt2Q4/GSs6u71UHKxB7S6JEMCp5Ve4xjh3eGQl/HRvg==" crossorigin="anonymous"> 29<link rel="stylesheet" type="text/css" href="https://fonts.googleapis.com/css?family=Roboto:300,400,500,700|Roboto+Slab:100,300,400,500,700|Material+Icons"> 30 31<!-- Code Syntax Highlighting --> 32<link rel="stylesheet" href="https://gitcdn.link/repo/jwarby/jekyll-pygments-themes/master/github.css" /> 33 34<!-- Styles 35 36<link rel="icon" href="data:image/svg+xml,<svg xmlns=%22http://www.w3.org/2000/svg%22 viewBox=%220 0 100 100%22><text y=%22.9em%22 font-size=%2290%22>/assets/img/logo.png</text></svg>"> 37 --> 38<link rel="stylesheet" href="/assets/css/main.css"> 39<link rel="canonical" href="/"> 40<link rel="shortcut icon" type="image/png" href="/assets/img/logo.png"/> 41 42<!-- JQuery --> 43<!-- jQuery -->
44<script src="https://cdnjs.cloudflare.com/ajax/libs/jquery/3.5.1/jquery.min.js" integrity="sha512-bLT0Qm9VnAYZDflyKcBaQ2gg0hSYNQrJ8RilYldYQ1FxQYoCLtUjuuRuZo+fjqhx/qtq/1itJ0C2ejDxltZVFg==" crossorigin="anonymous"></script>
44 45 46 47<!-- Theming--> 48
49<script src="/assets/js/theme.js"></script>
vendor: 1 bytes, line 49
49
50<script src="/assets/js/dark_mode.js"></script>
50 51 52 53 54<!-- Global site tag (gtag.js) - Google Analytics -->
vendor: 64 bytes, lines 54-55
54 55<script async src="https://www.googletagmanager.com/gtag/js?id=
55UA-XXXXXXXXX
vendor: 12 bytes, line 55
55"></script>
56<script> 57
vendor: 137 bytes, lines 57-61
57window.dataLayer = window.dataLayer || []; 58 function gtag() { dataLayer.push(arguments); } 59 gtag('js', new Date()); 60 61 gtag('config', '
61UA-XXXXXXXXX
vendor: 4 bytes, line 61
61');
62</script>
62 63 64 65 66 67 68<!-- MathJax -->
69<script type="text/javascript"> 70 window.MathJax = { 71 tex: { 72 tags: 'ams' 73 } 74 }; 75</script>
vendor: 1 bytes, line 75
75
76<script defer type="text/javascript" id="MathJax-script" src="https://cdn.jsdelivr.net/npm/[email protected]/es5/tex-mml-chtml.js"></script>
vendor: 1 bytes, line 76
76
77<script defer src="https://polyfill.io/v3/polyfill.min.js?features=es6"></script>
77 78 79 80 </head> 81 82 <body class="fixed-top-nav sticky-bottom-footer"> 83 84 <!-- Header --> 85 86 <header> 87 88 <!-- Nav Bar --> 89 <nav id="navbar" class="navbar navbar-light navbar-expand-sm fixed-top"> 90 <div class="container"> 91 92 <!-- Social Icons --> 93 <div class="navbar-brand social"> 94 <a href="mailto:%68%61%72%73%68%61%76%61%72%64%68%61%6E%38%36%34.%68%6B@%67%6D%61%69%6C.%63%6F%6D"><i class="fas fa-envelope"></i></a> 95 96<a href="https://scholar.google.com/citations?user=https://scholar.google.com/citations?user=LNXEjT8AAAAJ&hl=en" title="Google Scholar" target="_blank" rel="noopener noreferrer"><i class="ai ai-google-scholar"></i></a> 97 98 99<a href="https://github.com/kage08" title="GitHub" target="_blank" rel="noopener noreferrer"><i class="fab fa-github"></i></a> 100 101<a href="https://twitter.com/harsha_64" title="Twitter" target="_blank" rel="noopener noreferrer"><i class="fab fa-twitter"></i></a> 102 103 104 105 106 107 108 109 110 111 112 113<a href="https://kage08.github.io/feed.xml" title="RSS Feed"><i class="fas fa-rss-square"></i></a> 114 115 </div> 116 117 <!-- Navbar Toggle --> 118 <button class="navbar-toggler collapsed ml-auto" type="button" data-toggle="collapse" data-target="#navbarNav" aria-controls="navbarNav" aria-expanded="false" aria-label="Toggle navigation"> 119 <span class="sr-only">Toggle navigation</span> 120 <span class="icon-bar top-bar"></span> 121 <span class="icon-bar middle-bar"></span> 122 <span class="icon-bar bottom-bar"></span> 123 </button> 124 <div class="collapse navbar-collapse text-right" id="navbarNav"> 125 <ul class="navbar-nav ml-auto flex-nowrap"> 126 <!-- About --> 127 <li class="nav-item active"> 128 <a class="nav-link" href="/"> 129 about 130 131 <span class="sr-only">(current)</span> 132 133 </a> 134 </li> 135 136 <!-- Other pages --> 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 <li class="nav-item "> 158 <a class="nav-link" href="/publications/"> 159 publications 160 161 </a> 162 </li> 163 164 165 166 167 168 169 170 <!-- CV --> 171 <li class="nav-item "> 172 <a class="nav-link" target="_blank" href="/assets/pdf/Resume_new1.pdf">CV</a> 173 </li> 174 175 <div class="toggle-container"> 176 <a id="light-toggle"> 177 <i class="fas fa-moon"></i> 178 <i class="fas fa-sun"></i> 179 </a> 180 </div> 181 182 </ul> 183 </div> 184 </div> 185 </nav> 186 187</header> 188 189 190 <!-- Content --> 191 192 <div class="container mt-5"> 193 <div class="post"> 194 195 <header class="post-header"> 196 <h1 class="post-title"> 197 Harshavardhan Kamarthi 198 </h1> 199 <p class="desc"></p> 200 </header> 201 202 <article> 203 204 <div class="profile float-right"> 205 206 <img class="img-fluid z-depth-1 rounded" src="/assets/img/prof_pic.jpg"> 207 208 209 </div> 210 211 212 <div class="clearfix"> 213 <p>Iâm Harsha. Iâm a final-year <a href="https://ml.gatech.edu" target="_blank" rel="noopener noreferrer">Machine Learning PhD student</a> in the Department of Computational Science and Engineering at <a href="https://cc.gatech.edu/" target="_blank" rel="noopener noreferrer">Georgia Institute of Technology</a>. I am affiliated with AdityaLab and am advised by <a href="https://www.cc.gatech.edu/~badityap/" target="_blank" rel="noopener noreferrer">Dr. B Aditya Prakash</a>. I graduated from <a href="https://www.cse.iitm.ac.in" target="_blank" rel="noopener noreferrer">Indian Institute of Technology Madras</a> and am fortunate to have worked with <a href="http://www.cse.iitm.ac.in/~ravi/index.html" target="_blank" rel="noopener noreferrer">Dr. Balaraman Ravindran</a> and <a href="http://www.cse.iitm.ac.in/~sutanuc/" target="_blank" rel="noopener noreferrer">Dr. Sutanu Chakraborti</a>.</p> 214 215<p>My research broadly revolves around robust time-series forecasting and analysis, with a focus on uncertainty, scalability, cross-domain generalization, and operational deployment. My current research interests include:</p> 216 217<ol> 218 <li>Foundational time-series models that are pre-trained on multi-domain datasets and generalize across a wide range of domains and tasks (<a href="https://arxiv.org/abs/2311.11413" target="_blank" rel="noopener noreferrer">LPTM â24</a>
218, <a href="https://arxiv.org/abs/2311.07841" target="_blank" rel="noopener noreferrer">PEMS â23</a>, <a href="https://arxiv.org/pdf/2402.16132" target="_blank" rel="noopener noreferrer">LSTPrompt â24</a>, <a href="https://arxiv.org/abs/2406.08627" target="_blank" rel="noopener noreferrer">Time-MMD â24</a>, <a href="https://arxiv.org/abs/2602.20307" target="_blank" rel="noopener noreferrer">ICTP â25</a>, <a href="https://arxiv.org/abs/2503.11835" target="_blank" rel="noopener noreferrer">MM4TSA â25</a>).</li> 219 <li>Scalable and efficient time-series systems that handle large-scale operational and industrial data while providing robust and calibrated forecasts (<a href="https://github.com/AdityaLab/HAILS" target="_blank" rel="noopener noreferrer">HAILS â24</a>, <a href="https://arxiv.org/abs/2206.07940" target="_blank" rel="noopener noreferrer">ProfHiT â23</a>, <a href="https://arxiv.org/abs/2601.04432" target="_blank" rel="noopener noreferrer">AHA â26</a>).</li> 220 <li>Probabilistic and explainable time-series forecasting models that provide uncertainty estimates and remain robust to outliers, missing data, novel scenarios, and hierarchical constraints (<a href="https://arxiv.org/abs/2407.02641" target="_blank" rel="noopener noreferrer">STOIC â24</a>, <a href="https://arxiv.org/abs/2109.07438" target="_blank" rel="noopener noreferrer">CAMUL â22</a>, <a href="https://arxiv.org/abs/2106.03904" target="_blank" rel="noopener noreferrer">EPIFNP â21</a>, <a href="https://github.com/AdityaLab/Back2Future" target="_blank" rel="noopener noreferrer">B2F â22</a>, <a href="https://arxiv.org/abs/2603.06555" target="_blank" rel="noopener noreferrer">HiDeX â26</a>).</li> 221</ol> 222 223 </div> 224 225 226 <div class="news"> 227 <h2>news</h2> 228 229 <div class="table-responsive"> 230 <table class="table table-sm table-borderless"> 231 232 233 <tr> 234 <th scope="row">Mar 6, 2026</th> 235 <td> 236 237 <a href="https://arxiv.org/abs/2603.06555" target="_blank" rel="noopener noreferrer"><em>Hierarchical Industrial Demand Forecasting with Temporal and Uncertainty Explanations</em></a> is accepted at <strong>ICDE 2026</strong>! 238 239 240 </td> 241 </tr> 242 243 <tr> 244 <th scope="row">Jan 7, 2026</th> 245 <td> 246 247 <a href="https://arxiv.org/abs/2601.04432" target="_blank" rel="noopener noreferrer"><em>AHA: Scalable Alternative History Analysis for Operational Timeseries Applications</em></a> will appear at <strong>KDD 2026</strong>! 248 249 250 </td> 251 </tr> 252 253 <tr> 254 <th scope="row">Oct 21, 2025</th> 255 <td> 256 257 <a href="https://arxiv.org/abs/2602.20307" target="_blank" rel="noopener noreferrer"><em>In-context Pre-trained Time-Series Foundation Models adapt to Unseen Tasks</em></a> appeared at <strong>CIKM 2025</strong>! 258 259 260 </td> 261 </tr> 262 263 <tr> 264 <th scope="row">Mar 15, 2025</th> 265 <td> 266 267 <a href="https://github.com/AdityaLab/Samay" target="_blank" rel="noopener noreferrer"><em>Samay</em></a> a easy to use library for time-series foundational models is released! Do check it out! 268 269 270 </td> 271 </tr> 272 273 <tr> 274 <th scope="row">Mar 14, 2025</th> 275 <td> 276 277 Our survey <a href="https://arxiv.org/abs/2503.11835" target="_blank" rel="noopener noreferrer"><em>How Can Time Series Analysis Benefit From Multiple Modalities?</em></a> is now available on arXiv. 278 279 280 </td> 281 </tr> 282 283 </table> 284 </div> 285 286</div> 287 288 289 290 291 <div class="publications"> 292 <h2>publications</h2> 293 <ol class="bibliography"> 294<li> 295<div class="row"> 296 <div class="col-sm-2 abbr"> 297 298 <img class="img-fluid" src="/assets/pubimg/hidex26.png" alt="hidex26"> 299 300 </div> 301 302 <div id="kamarthi2026hidex" class="col-sm-8"> 303 304 <div class="title">Hierarchical Industrial Demand Forecasting with Temporal and Uncertainty Explanations</div> 305 <div class="author"> 306 307 308 309 310 311 312 313 314 315 <em>Kamarthi, Harshavardhan</em>, 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 <a href="https://scholar.google.com/citations?user=LxyfYAUAAAAJ&hl=zh-CN" target="_blank" rel="noopener noreferrer">Xu, Shangqing</a>, 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 Tong, Xinjie, 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 Zhou, Xingyu, 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 Peters, James, 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 Czyzyk, Joseph, 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 415 416 417 418 419 420 </div> 421 422 <div class="periodical"> 423 424 <em>ICDE</em> 425 426 427 2026 428 429 </div> 430 431 432 <div class="links"> 433 434 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 435 436 437 438 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 439 440 441 442 443 <a href="https://arxiv.org/abs/2603.06555" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 444 445 446 447 448 449 450 451 452 </div> 453 454 <!-- Hidden abstract block --> 455 456 <div class="abstract hidden"> 457 <p>Hierarchical time-series forecasting is essential for demand prediction across industries, but the interpretability of such forecasts remains largely unexplored. We introduce an interpretability method for large hierarchical probabilistic time-series forecasting that adapts generic interpretability techniques while addressing hierarchy and uncertainty. The approach explains the significance of time series and external variables at specific time points, the impact of variables on forecast uncertainty, and forecast changes caused by training-data modifications. Experiments on semi-synthetic datasets based on industrial demand scenarios and real-world case studies demonstrate that the method explains state-of-the-art industrial forecasting models with significantly higher explainability accuracy.</p> 458 </div> 459 460 461 <!-- Hidden bibtex block --> 462 463 <div class="bibtex hidden"> 464 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">kamarthi2026hidex</span><span class="p">,</span> 465 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{hidex26}</span><span class="p">,</span> 466 <span class="na">title</span> <span class="p">=</span> <span class="s">{Hierarchical Industrial Demand Forecasting with Temporal and Uncertainty Explanations}</span><span class="p">,</span> 467 <span class="na">author</span> <span class="p">=</span> <span class="s">{Kamarthi, Harshavardhan and Xu, Shangqing and Tong, Xinjie and Zhou, Xingyu and Peters, James and Czyzyk, Joseph and Prakash, B Aditya}</span><span class="p">,</span> 468 <span class="na">
468journal</span> <span class="p">=</span> <span class="s">{ICDE}</span><span class="p">,</span> 469 <span class="na">year</span> <span class="p">=</span> <span class="s">{2026}</span><span class="p">,</span> 470 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 471 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2603.06555}</span><span class="p">,</span> 472 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span> 473<span class="p">}</span></code></pre></figure> 474 </div> 475 476 </div> 477</div> 478</li> 479<li> 480<div class="row"> 481 <div class="col-sm-2 abbr"> 482 483 <img class="img-fluid" src="/assets/pubimg/aha26.png" alt="aha26"> 484 485 </div> 486 487 <div id="kamarthi2026aha" class="col-sm-8"> 488 489 <div class="title">AHA: Scalable Alternative History Analysis for Operational Timeseries Applications</div> 490 <div class="author"> 491 492 493 494 495 496 497 498 499 500 <em>Kamarthi, Harshavardhan</em>, 501 502 503 504 505 506 507 508 509 510 511 512 513 514 Shah, Harshil, 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 Milner, Henry, 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 Sinha, Sayan, 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 Li, Yan, 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a>, 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 and Sekar, Vyas 595 596 597 598 599 600 </div> 601 602 <div class="periodical"> 603 604 <em>To appear in KDD</em> 605 606 607 2026 608 609 </div> 610 611 612 <div class="links"> 613 614 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 615 616 617 618 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 619 620 621 622 623 <a href="https://arxiv.org/abs/2601.04432" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 624 625 626 627 628 629 630 631 632 </div> 633 634 <!-- Hidden abstract block --> 635 636 <div class="abstract hidden"> 637 <p>Many operational systems collect high-dimensional timeseries data about users or systems on key performance metrics. Over such historical data, operators and data analysts often need to run retrospective analysis, such as analyzing anomaly detection algorithms, experimenting with different alert configurations, and evaluating new algorithms. We refer to this class of workloads as alternative history analysis for operational datasets. We design and implement AHA, a system that provides cost efficiency and fidelity for high-dimensional data by leveraging decomposable statistics, sparsity in active subpopulations, and efficient aggregation structure in modern analytics databases. Using multiple real-world datasets and production-pipeline case studies at a large video analytics company, AHA provides 100% accuracy for a broad range of downstream tasks and up to 85x lower total cost of ownership compared to conventional methods.</p> 638 </div> 639 640 641 <!-- Hidden bibtex block --> 642 643 <div class="bibtex hidden"> 644 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">kamarthi2026aha</span><span class="p">,</span> 645 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{aha26}</span><span class="p">,</span> 646 <span class="na">title</span> <span class="p">=</span> <span class="s">{AHA: Scalable Alternative History Analysis for Operational Timeseries Applications}</span><span class="p">,</span> 647 <span class="na">author</span> <span class="p">=</span> <span class="s">{Kamarthi, Harshavardhan and Shah, Harshil and Milner, Henry and Sinha, Sayan and Li, Yan and Prakash, B Aditya and Sekar,
647Vyas}</span><span class="p">,</span> 648 <span class="na">journal</span> <span class="p">=</span> <span class="s">{To appear in KDD}</span><span class="p">,</span> 649 <span class="na">year</span> <span class="p">=</span> <span class="s">{2026}</span><span class="p">,</span> 650 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 651 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2601.04432}</span><span class="p">,</span> 652 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span> 653<span class="p">}</span></code></pre></figure> 654 </div> 655 656 </div> 657</div> 658</li> 659<li> 660<div class="row"> 661 <div class="col-sm-2 abbr"> 662 663 <img class="img-fluid" src="/assets/pubimg/mm4tsa25.png" alt="mm4tsa25"> 664 665 </div> 666 667 <div id="liu2025mm4tsa" class="col-sm-8"> 668 669 <div class="title">How Can Time Series Analysis Benefit From Multiple Modalities? A Survey and Outlook</div> 670 <div class="author"> 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 <a href="https://twitter.com/HessianLiu" target="_blank" rel="noopener noreferrer">Liu, Haoxin</a>, 687 688 689 690 691 692 693 694 695 696 697 698 699 700 <em>Kamarthi, Harshavardhan</em>, 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 <a href="https://twitter.com/leozhao_zhiyuan" target="_blank" rel="noopener noreferrer">Zhao, Zhiyuan</a>, 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 <a href="https://scholar.google.com/citations?user=LxyfYAUAAAAJ&hl=zh-CN" target="_blank" rel="noopener noreferrer">Xu, Shangqing</a>, 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 Wang, Shiyu, 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 Wen, Qingsong, 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 Hartvigsen, Tom, 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 Wang, Fei, 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 820 821 822 823 824 825 </div> 826 827 <div class="periodical"> 828 829 <em>arXiv preprint arXiv:2503.11835</em> 830 831 832 2025 833 834 </div> 835 836 837 <div class="links"> 838 839 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 840 841 842 843 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 844 845 846 847 848 <a href="https://arxiv.org/abs/2503.11835" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 849 850 851 852 853 854 <a href="https://github.com/qingsongedu/Awesome-TimeSeries-SpatioTemporal-LM-LLM" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">
854Code</a> 855 856 857 858 859 </div> 860 861 <!-- Hidden abstract block --> 862 863 <div class="abstract hidden"> 864 <p>Time series analysis is a longstanding research topic with wide real-world significance, but compared with language and vision it remains relatively underexplored and isolated. This survey reviews Multiple Modalities for Time Series Analysis (MM4TSA), an emerging area studying how time series analysis can benefit from richer modalities. We systematically discuss three benefits: reusing foundation models from other modalities for efficient time series analysis, multimodal extension for enhanced time series analysis, and cross-modality interaction for advanced time series analysis. We group works by modality type, including text, images, audio, tables, and others, and identify future opportunities in modality selection, heterogeneous modality combinations, and unseen-task generalization.</p> 865 </div> 866 867 868 <!-- Hidden bibtex block --> 869 870 <div class="bibtex hidden"> 871 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">liu2025mm4tsa</span><span class="p">,</span> 872 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{mm4tsa25}</span><span class="p">,</span> 873 <span class="na">title</span> <span class="p">=</span> <span class="s">{How Can Time Series Analysis Benefit From Multiple Modalities? A Survey and Outlook}</span><span class="p">,</span> 874 <span class="na">author</span> <span class="p">=</span> <span class="s">{Liu, Haoxin and Kamarthi, Harshavardhan and Zhao, Zhiyuan and Xu, Shangqing and Wang, Shiyu and Wen, Qingsong and Hartvigsen, Tom and Wang, Fei and Prakash, B Aditya}</span><span class="p">,</span> 875 <span class="na">journal</span> <span class="p">=</span> <span class="s">{arXiv preprint arXiv:2503.11835}</span><span class="p">,</span> 876 <span class="na">year</span> <span class="p">=</span> <span class="s">{2025}</span><span class="p">,</span> 877 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 878 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2503.11835}</span><span class="p">,</span> 879 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/qingsongedu/Awesome-TimeSeries-SpatioTemporal-LM-LLM}</span><span class="p">,</span> 880 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span> 881<span class="p">}</span></code></pre></figure> 882 </div> 883 884 </div> 885</div> 886</li> 887<li> 888<div class="row"> 889 <div class="col-sm-2 abbr"> 890 891 <img class="img-fluid" src="/assets/pubimg/ictp25.png" alt="ictp25"> 892 893 </div> 894 895 <div id="xu2025ictp" class="col-sm-8"> 896 897 <div class="title">In-context Pre-trained Time-Series Foundation Models adapt to Unseen Tasks</div> 898 <div class="author"> 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 <a href="https://scholar.google.com/citations?user=LxyfYAUAAAAJ&hl=zh-CN" target="_blank" rel="noopener noreferrer">Xu, Shangqing</a>, 915 916 917 918 919 920 921 922 923 924 925 926 927 928 <em>Kamarthi, Harshavardhan</em>, 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 <a href="https://twitter.com/HessianLiu" target="_blank" rel="noopener noreferrer">Liu, Haoxin</a>, 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 968 969 970 971 972 973 </div> 974 975 <div class="periodical"> 976 977 <em>In CIKM</em> 978 979 980 2025 981 982 </div> 983 984 985 <div class="links"> 986 987 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 988 989 990 991 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 992 993 994 995 996 <a href="https://arxiv.org/abs/2602.20307" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 997 998 999 1000 1001 1002 1003 1004 1005 </div> 1006 1007 <!-- Hidden abstract block --> 1008 1009 <div class="abstract hidden"> 1010 <p>Time-series foundation models have demonstrated strong generalization across diverse datasets and tasks, but existing models are usually pre-trained for specific tasks and often struggle to generalize to unseen tasks without fine-tuning. We propose In-Context Time-series Pre-training (ICTP), which restructures pre-training data to equip a backbone time-series foundation model with in-context learning capabilities. ICTP enables test-time adaptation to unseen tasks from input-output relationships provided in context and improves state-of-the-art time-series foundation models by approximately 11.4% on unseen tasks without fine-tuning.</p> 1011 </div> 1012 1013 1014 <!-- Hidden bibtex block --> 1015 1016 <div class="bibtex hidden"> 1017 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">xu2025ictp</span><span class="p">,</span> 1018 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{ictp25}</span><span class="p">,</span> 1019 <span class="na">title</span> <span class="p">=</span> <span class="s">{In-context Pre-trained Time-Series Foundation Models adapt to Unseen Tasks}</span><span class="p">,</span> 1020 <span class="na">author</span> <span class="p">=</span> <span class="s">{Xu, Shangqing and Kamarthi, Harshavardhan and Liu, Haoxin and Prakash, B Aditya}</span><span class="p">,</span> 1021 <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{CIKM}</span><span class="p">,</span> 1022 <span class="na">pages</span> <span class="p">=</span> <span class="s">{5386--5390}</span><span class="p">,</span> 1023 <span class="na">year</span> <span class="p">=</span> <span class="s">{2025}</span><span class="p">,</span> 1024 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 1025 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2602.20307}</span><span class="p">,</span> 1026 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span> 1027<span class="p">}</span></code></pre></figure> 1028 </div> 1029 1030 </div> 1031</div> 1032</li> 1033<li> 1034<div class="row"> 1035 <div class="col-sm-2 abbr"> 1036 1037 <img class="img-fluid" src="/assets/pubimg/lptm23.png" alt="lptm23"> 1038 1039 </div> 1040 1041 <div id="kamarthi2023large" class="col-sm-8"> 1042 1043 <div class="title">Large Pre-trained time series models for cross-domain Time series anal
1043ysis tasks</div> 1044 <div class="author"> 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 <em>Kamarthi, Harshavardhan</em>, 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 1074 1075 1076 1077 1078 1079 </div> 1080 1081 <div class="periodical"> 1082 1083 <em>NeurIPS</em> 1084 1085 1086 2024 1087 1088 </div> 1089 1090 1091 <div class="links"> 1092 1093 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 1094 1095 1096 1097 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 1098 1099 1100 1101 1102 <a href="https://arxiv.org/abs/2311.11413" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 1103 1104 1105 1106 1107 1108 <a href="https://github.com/kage08/SegmentTS" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">Code</a> 1109 1110 1111 1112 1113 </div> 1114 1115 <!-- Hidden abstract block --> 1116 1117 <div class="abstract hidden"> 1118 <p>Large pre-trained models have been vital in recent advancements in domains like language and vision, making model training for individual downstream tasks more efficient and provide superior performance. However, tackling time-series analysis tasks usually involves designing and training a separate model from scratch leveraging training data and domain expertise specific to the task. We tackle a significant challenge for pre-training a foundational time-series model from multi-domain time-series datasets: extracting 1119 semantically useful tokenized inputs to the model 1120 across heterogenous time-series from different domains. We propose Large Pre-trained Time-series Models (LPTM) that introduces a novel method of adaptive segmentation that automatically identifies optimal dataset-specific 1121 segmentation strategy during pre-training. This enables 1122 LPTM to perform similar to or better than domain-specific state-of-art model 1123 when fine-tuned to different downstream time-series analysis tasks and under zero-shot settings. 1124 LPTM achieves superior forecasting and time-series classification results 1125 taking up to 40% less data and 50% less training time 1126 compared to state-of-art baselines.</p> 1127 </div> 1128 1129 1130 <!-- Hidden bibtex block --> 1131 1132 <div class="bibtex hidden"> 1133 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">kamarthi2023large</span><span class="p">,</span> 1134 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{lptm23}</span><span class="p">,</span> 1135 <span class="na">title</span> <span class="p">=</span> <span class="s">{Large Pre-trained time series models for cross-domain Time series analysis tasks}</span><span class="p">,</span> 1136 <span class="na">author</span> <span class="p">=</span> <span class="s">{Kamarthi, Harshavardhan and Prakash, B Aditya}</span><span class="p">,</span> 1137 <span class="na">journal</span> <span class="p">=</span> <span class="s">{NeurIPS}</span><span class="p">,</span> 1138 <span class="na">year</span> <span class="p">=</span> <span class="s">{2024}</span><span class="p">,</span> 1139 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 1140 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2311.11413}</span><span class="p">,</span> 1141 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 1142 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/kage08/SegmentTS}</span> 1143<span class="p">}</span></code></pre></figure> 1144 </div> 1145 1146 </div> 1147</div> 1148</li> 1149<li> 1150<div class="row"> 1151 <div class="col-sm-2 abbr"> 1152 1153 <img class="img-fluid" src="/assets/pubimg/hails24.png" alt="hails24"> 1154 1155 </div> 1156 1157 <div id="kamarthi2024hails" class="col-sm-8"> 1158 1159 <div class="title">
1159Large Scale Hierarchical Industrial Demand Time-Series Forecasting incorporating Sparsity</div> 1160 <div class="author"> 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 <em>Kamarthi, Harshavardhan</em>, 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 Sasanur, Aditya B, 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 Tong, Xinjie, 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 Zhou, Xingyu, 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 Peters, James, 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 Czyzyk, Joe, 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 1265 1266 1267 1268 1269 1270 </div> 1271 1272 <div class="periodical"> 1273 1274 <em>KDD</em> 1275 1276 1277 2024 1278 1279 </div> 1280 1281 1282 <div class="links"> 1283 1284 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 1285 1286 1287 1288 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 1289 1290 1291 1292 1293 <a href="https://arxiv.org/abs/2407.02657" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 1294 1295 1296 1297 1298 1299 <a href="https://github.com/AdityaLab/HAILS" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">Code</a> 1300 1301 1302 1303 1304 </div> 1305 1306 <!-- Hidden abstract block --> 1307 1308 <div class="abstract hidden"> 1309 <p>Hierarchical time-series forecasting (HTSF) is an important problem for many real-world business applications where the goal is to simultaneously forecast multiple time-series that are related to each other via a hierarchical relation. Recent works, however, do not address two important challenges that are typically observed in many demand forecasting applications at large companies. First, many time-series at lower levels of the hierarchy have high sparsity i.e., they have a significant number of zeros. Most HTSF methods do not address this varying sparsity across the hierarchy. Further, they do not scale well to the large size of the real-world hierarchy typically unseen in benchmarks used in literature. We resolve both these challenges by proposing HAILS, a novel probabilistic hierarchical model that enables accurate and calibrated probabilistic forecasts across the hierarchy by adaptively modeling sparse and dense time-series with different distributional assumptions and reconciling them to adhere to hierarchical constraints. We show the scalability and effectiveness of our methods by evaluating them against real-world demand forecasting datasets. We deploy HAILS at a large chemical manufacturing company for a product demand forecasting application with over ten thousand products and observe a significant 8.5% improvement in forecast accuracy and 23% better improvement for sparse time-series. The enhanced accuracy and scalability make HAILS a valuable tool for improved business planning and customer experience.</p> 1310 </div> 1311 1312 1313 <!-- Hidden bibtex block --> 1314 1315 <div class="bibtex hidden"> 1316 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">kamarthi2024hails</span><span class="p">,</span> 1317 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{hails24}</span><span class="p">,</span> 1318 <span class="na">title</span> <span class="p">=</span> <span class="s">
1318{Large Scale Hierarchical Industrial Demand Time-Series Forecasting incorporating Sparsity}</span><span class="p">,</span> 1319 <span class="na">author</span> <span class="p">=</span> <span class="s">{Kamarthi, Harshavardhan and Sasanur, Aditya B and Tong, Xinjie and Zhou, Xingyu and Peters, James and Czyzyk, Joe and Prakash, B Aditya}</span><span class="p">,</span> 1320 <span class="na">journal</span> <span class="p">=</span> <span class="s">{KDD}</span><span class="p">,</span> 1321 <span class="na">pages</span> <span class="p">=</span> <span class="s">{5230--5239}</span><span class="p">,</span> 1322 <span class="na">year</span> <span class="p">=</span> <span class="s">{2024}</span><span class="p">,</span> 1323 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 1324 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2407.02657}</span><span class="p">,</span> 1325 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 1326 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/AdityaLab/HAILS}</span> 1327<span class="p">}</span></code></pre></figure> 1328 </div> 1329 1330 </div> 1331</div> 1332</li> 1333<li> 1334<div class="row"> 1335 <div class="col-sm-2 abbr"> 1336 1337 <img class="img-fluid" src="/assets/pubimg/foil24.png" alt="foil24"> 1338 1339 </div> 1340 1341 <div id="haoxin2024foil" class="col-sm-8"> 1342 1343 <div class="title">Time-Series Forecasting for Out-of-Distribution Generalization Using Invariant Learning</div> 1344 <div class="author"> 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 <a href="https://twitter.com/HessianLiu" target="_blank" rel="noopener noreferrer">Liu, Haoxin</a>, 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 <em>Kamarthi, Harshavardhan</em>, 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 <a href="https://lingkai-kong.com" target="_blank" rel="noopener noreferrer">Kong, Lingkai</a>, 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 <a href="https://twitter.com/leozhao_zhiyuan" target="_blank" rel="noopener noreferrer">Zhao, Zhiyuan</a>, 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 <a href="https://chaozhang.org" target="_blank" rel="noopener noreferrer">Zhang, Chao</a>, 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 1454 1455 1456 1457 1458 1459 </div> 1460 1461 <div class="periodical"> 1462 1463 <em>In ICML</em> 1464 1465 1466 2024 1467 1468 </div> 1469 1470 1471 <div class="links"> 1472 1473 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 1474 1475 1476 1477 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 1478 1479 1480 1481 1482 <a href="https://openreview.net/forum?id=SMUXPVKUBg" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 1483 1484 1485 1486 1487 1488 <a href="https://github.com/AdityaLab/FOIL" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">
1488Code</a> 1489 1490 1491 1492 1493 </div> 1494 1495 <!-- Hidden abstract block --> 1496 1497 <div class="abstract hidden"> 1498 <p>Time-series forecasting (TSF) finds broad applications in real-world scenarios. Due to the dynamic nature of time-series data, it is crucial for TSF models to preserve out-of-distribution (OOD) generalization abilities, as training and test sets represent historical and future data respectively. In this paper, we aim to alleviate the inherent OOD problem in TSF via invariant learning. We identify fundamental challenges of invariant learning for TSF. First, the target variables in TSF may not be sufficiently determined by the input due to unobserved core variables in TSF, breaking the fundamental assumption of invariant learning. Second, time-series datasets lack adequate environment labels, while existing environmental inference methods are not suitable for TSF. To address these challenges, we propose FOIL, a model-agnostic framework that endows time-series forecasting for out-of-distribution generalization via invariant learning. Specifically, FOIL employs a novel surrogate loss to mitigate the impact of unobserved variables. Further, FOIL implements joint optimization by alternately inferring environments effectively with a multi-head network while preserving the temporal adjacency structure and learning invariant representations across inferred environments for OOD generalized TSF. Extensive experiments demonstrate that the proposed FOIL significantly and consistently improves the performance of various TSF models, achieving gains of up to 85%.</p> 1499 </div> 1500 1501 1502 <!-- Hidden bibtex block --> 1503 1504 <div class="bibtex hidden"> 1505 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">haoxin2024foil</span><span class="p">,</span> 1506 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{foil24}</span><span class="p">,</span> 1507 <span class="na">title</span> <span class="p">=</span> <span class="s">{Time-Series Forecasting for Out-of-Distribution Generalization Using Invariant Learning}</span><span class="p">,</span> 1508 <span class="na">author</span> <span class="p">=</span> <span class="s">{Liu, Haoxin and Kamarthi, Harshavardhan and Kong, Lingkai and Zhao, Zhiyuan and Zhang, Chao and Prakash, B Aditya}</span><span class="p">,</span> 1509 <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{ICML}</span><span class="p">,</span> 1510 <span class="na">pages</span> <span class="p">=</span> <span class="s">{31312--31325}</span><span class="p">,</span> 1511 <span class="na">year</span> <span class="p">=</span> <span class="s">{2024}</span><span class="p">,</span> 1512 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 1513 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://openreview.net/forum?id=SMUXPVKUBg}</span><span class="p">,</span> 1514 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 1515 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/AdityaLab/FOIL}</span> 1516<span class="p">}</span></code></pre></figure> 1517 </div> 1518 1519 </div> 1520</div> 1521</li> 1522<li> 1523<div class="row"> 1524 <div class="col-sm-2 abbr"> 1525 1526 <img class="img-fluid" src="/assets/pubimg/lstprompt24.png" alt="lstprompt24"> 1527 1528 </div> 1529 1530 <div id="liu2024lstprompt" class="col-sm-8"> 1531 1532 <div class="title">Lstprompt: Large language models as zero-shot time series forecasters by long-short-term prompting</div> 1533 <div class="author"> 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 <a href="https://twitter.com/HessianLiu" target="_blank" rel="noopener noreferrer">Liu, Haoxin</a>, 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 <a href="https://twitter.com/leozhao_zhiyuan" target="_blank" rel="noopener noreferrer">Zhao, Zhiyuan</a>, 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 Wang, Jindong, 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 <em>Kamarthi, Harshavardhan</em>, 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 1618 1619 1620 1621 1622 1623 </div> 1624 1625 <div class="periodical"> 1626 1627 <em>ACL Findings</em> 1628 1629 1630 2024 1631 1632 </div> 1633 1634 1635 <div class="links"> 1636 1637 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 1638 1639 1640 1641 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 1642 1643 1644 1645 1646 <a href="https://arxiv.org/pdf/2402.16132" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 1647 1648 1649 1650 1651 1652 <a href="https://github.com/AdityaLab/lstprompt" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">
1652Code</a> 1653 1654 1655 1656 1657 </div> 1658 1659 <!-- Hidden abstract block --> 1660 1661 <div class="abstract hidden"> 1662 <p>Time-series forecasting (TSF) finds broad applications in real-world scenarios. Prompting off-the-shelf Large Language Models (LLMs) demonstrates strong zero-shot TSF capabilities while preserving computational efficiency. However, existing prompting methods oversimplify TSF as language next-token predictions, overlooking its dynamic nature and lack of integration with state-of-the-art prompt strategies such as Chain-of-Thought. Thus, we propose LSTPrompt, a novel approach for prompting LLMs in zero-shot TSF tasks. LSTPrompt decomposes TSF into short-term and long-term forecasting sub-tasks, tailoring prompts to each. LSTPrompt guides LLMs to regularly reassess forecasting mechanisms to enhance adaptability. Extensive evaluations demonstrate consistently better performance of LSTPrompt than existing prompting methods, and competitive results compared to foundation TSF models.</p> 1663 </div> 1664 1665 1666 <!-- Hidden bibtex block --> 1667 1668 <div class="bibtex hidden"> 1669 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">liu2024lstprompt</span><span class="p">,</span> 1670 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{lstprompt24}</span><span class="p">,</span> 1671 <span class="na">title</span> <span class="p">=</span> <span class="s">{Lstprompt: Large language models as zero-shot time series forecasters by long-short-term prompting}</span><span class="p">,</span> 1672 <span class="na">author</span> <span class="p">=</span> <span class="s">{Liu, Haoxin and Zhao, Zhiyuan and Wang, Jindong and Kamarthi, Harshavardhan and Prakash, B Aditya}</span><span class="p">,</span> 1673 <span class="na">journal</span> <span class="p">=</span> <span class="s">{ACL Findings}</span><span class="p">,</span> 1674 <span class="na">pages</span> <span class="p">=</span> <span class="s">{7832--7840}</span><span class="p">,</span> 1675 <span class="na">year</span> <span class="p">=</span> <span class="s">{2024}</span><span class="p">,</span> 1676 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 1677 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/pdf/2402.16132}</span><span class="p">,</span> 1678 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 1679 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/AdityaLab/lstprompt}</span> 1680<span class="p">}</span></code></pre></figure> 1681 </div> 1682 1683 </div> 1684</div> 1685</li> 1686<li> 1687<div class="row"> 1688 <div class="col-sm-2 abbr"> 1689 1690 <img class="img-fluid" src="/assets/pubimg/timemmd24.png" alt="timemmd24"> 1691 1692 </div> 1693 1694 <div id="liu2024time" class="col-sm-8"> 1695 1696 <div class="title">Time-MMD: Multi-Domain Multimodal Dataset for Time Series Analysis</div> 1697 <div class="author"> 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 <a href="https://twitter.com/HessianLiu" target="_blank" rel="noopener noreferrer">Liu, Haoxin</a>, 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 <a href="https://scholar.google.com/citations?user=LxyfYAUAAAAJ&hl=zh-CN" target="_blank" rel="noopener noreferrer">Xu, Shangqing</a>, 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 <a href="https://twitter.com/leozhao_zhiyuan" target="_blank" rel="noopener noreferrer">Zhao, Zhiyuan</a>, 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 <a href="https://lingkai-kong.com" target="_blank" rel="noopener noreferrer">Kong, Lingkai</a>, 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 <em>Kamarthi, Harshavardhan</em>, 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 Sasanur, Aditya B, 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 Sharma, Megha, 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 Cui, Jiaming, 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 Wen, Qingsong, 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 <a href="https://chaozhang.org" target="_blank" rel="noopener noreferrer">Zhang, Chao</a>, 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 and others, 1882 1883 1884 1885 1886 1887 </div> 1888 1889 <div class="periodical"> 1890 1891 <em>NeurIPS</em> 1892 1893 1894 2024 1895 1896 </div> 1897 1898 1899 <div class="links"> 1900 1901 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 1902 1903 1904 1905 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 1906 1907 1908 1909 1910 <a href="https://arxiv.org/abs/2406.08627" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 1911 1912 1913 1914 1915 1916 <a href="https://github.com/AdityaLab/Time-MMD" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">
1916Code</a> 1917 1918 1919 1920 1921 </div> 1922 1923 <!-- Hidden abstract block --> 1924 1925 <div class="abstract hidden"> 1926 <p>Time series data are ubiquitous across a wide range of real-world domains. While real-world time series analysis (TSA) requires human experts to integrate numerical series data with multimodal domain-specific knowledge, most existing TSA models rely solely on numerical data, overlooking the significance of information beyond numerical series. This oversight is due to the untapped potential of textual series data and the absence of a comprehensive, high-quality multimodal dataset. To overcome this obstacle, we introduce Time-MMD, the first multi-domain, multimodal time series dataset covering 9 primary data domains. Time-MMD ensures fine-grained modality alignment, eliminates data contamination, and provides high usability. Additionally, we develop MM-TSFlib, the first multimodal time-series forecasting (TSF) library, seamlessly pipelining multimodal TSF evaluations based on Time-MMD for in-depth analyses. Extensive experiments conducted on Time-MMD through MM-TSFlib demonstrate significant performance enhancements by extending unimodal TSF to multimodality, evidenced by over 15% mean squared error reduction in general, and up to 40% in domains with rich textual data. More importantly, our datasets and library revolutionize broader applications, impacts, research topics to advance TSA.</p> 1927 </div> 1928 1929 1930 <!-- Hidden bibtex block --> 1931 1932 <div class="bibtex hidden"> 1933 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">liu2024time</span><span class="p">,</span> 1934 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{timemmd24}</span><span class="p">,</span> 1935 <span class="na">title</span> <span class="p">=</span> <span class="s">{Time-MMD: Multi-Domain Multimodal Dataset for Time Series Analysis}</span><span class="p">,</span> 1936 <span class="na">author</span> <span class="p">=</span> <span class="s">{Liu, Haoxin and Xu, Shangqing and Zhao, Zhiyuan and Kong, Lingkai and Kamarthi, Harshavardhan and Sasanur, Aditya B and Sharma, Megha and Cui, Jiaming and Wen, Qingsong and Zhang, Chao and others}</span><span class="p">,</span> 1937 <span class="na">journal</span> <span class="p">=</span> <span class="s">{NeurIPS}</span><span class="p">,</span> 1938 <span class="na">year</span> <span class="p">=</span> <span class="s">{2024}</span><span class="p">,</span> 1939 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 1940 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2406.08627}</span><span class="p">,</span> 1941 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 1942 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/AdityaLab/Time-MMD}</span> 1943<span class="p">}</span></code></pre></figure> 1944 </div> 1945 1946 </div> 1947</div> 1948</li> 1949<li> 1950<div class="row"> 1951 <div class="col-sm-2 abbr"> 1952 1953 <img class="img-fluid" src="/assets/pubimg/nature24.png" alt="nature24"> 1954 1955 </div> 1956 1957 <div id="mathis2024title" class="col-sm-8"> 1958 1959 <div class="title">Title evaluation of FluSight influenza forecasting in the 2021â22 and 2022â23 seasons with a new target laboratory-confirmed influenza hospitalizations</div> 1960 <div class="author"> 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 Mathis, Sarabeth M, 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 Webber, Alexander E, 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 León, Tomás M, 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Murray, Erin L, 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 Sun, Monica, 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 White, Lauren A, 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 Brooks, Logan C, 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 Green, Alden, 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 Hu, Addison J, 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 Rosenfeld, Roni, 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 and others, 2122 2123 2124 2125 2126 2127 </div> 2128 2129 <div class="periodical"> 2130 2131 <em>Nature Communications</em> 2132 2133 2134 2024 2135 2136 </div> 2137 2138 2139 <div class="links"> 2140 2141 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 2142 2143 2144 2145 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 2146 2147 2148 2149 2150 <a href="https://www.nature.com/articles/s41467-024-50601-9" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 2151 2152 2153 2154 2155 2156 2157 2158 2159 </div> 2160 2161 <!-- Hidden abstract block --> 2162 2163 <div class="abstract hidden"> 2164 <p>Accurate forecasts can enable more effective public health responses during seasonal influenza epidemics. For the 2021â22 and 2022â23 influenza seasons, 26 forecasting teams provided national and jurisdiction-specific probabilistic predictions of weekly confirmed influenza hospital admissions for one-to-four weeks ahead. Forecast skill is evaluated using the Weighted Interval Score (WIS), relative WIS, and coverage. Six out of 23 models outperform the baseline model across forecast weeks and locations in 2021â22 and 12 out of 18 models in 2022â23. Averaging across all forecast targets, the FluSight ensemble is the 2nd most accurate model measured by WIS in 2021â22 and the 5th most accurate in the 2022â23 season. Forecast skill and 95% coverage for the FluSight ensemble and most component models degrade over longer forecast horizons. In this work we demonstrate that while the FluSight ensemble was a robust predictor, even ensembles face challenges during periods of rapid change.</p> 2165 </div> 2166 2167 2168 <!-- Hidden bibtex block --> 2169 2170 <div class="bibtex hidden"> 2171 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">mathis2024title</span><span class="p">,</span> 2172 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{nature24}</span><span class="p">,</span> 2173 <span class="na">title</span> <span class="p">=</span> <span class="s">{Title evaluation of FluSight influenza forecasting in the 2021--22 and 2022--23 seasons with a new target laboratory-confirmed influenza hospital
2173izations}</span><span class="p">,</span> 2174 <span class="na">author</span> <span class="p">=</span> <span class="s">{Mathis, Sarabeth M and Webber, Alexander E and Le{\'o}n, Tom{\'a}s M and Murray, Erin L and Sun, Monica and White, Lauren A and Brooks, Logan C and Green, Alden and Hu, Addison J and Rosenfeld, Roni and others}</span><span class="p">,</span> 2175 <span class="na">journal</span> <span class="p">=</span> <span class="s">{Nature Communications}</span><span class="p">,</span> 2176 <span class="na">year</span> <span class="p">=</span> <span class="s">{2024}</span><span class="p">,</span> 2177 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 2178 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://www.nature.com/articles/s41467-024-50601-9}</span><span class="p">,</span> 2179 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span> 2180<span class="p">}</span></code></pre></figure> 2181 </div> 2182 2183 </div> 2184</div> 2185</li> 2186<li> 2187<div class="row"> 2188 <div class="col-sm-2 abbr"> 2189 2190 <img class="img-fluid" src="/assets/pubimg/stoic24.png" alt="stoic24"> 2191 2192 </div> 2193 2194 <div id="kamarthi2024stoic" class="col-sm-8"> 2195 2196 <div class="title">Learning Graph Structures and Uncertainty for Accurate and Calibrated Time-series Forecasting</div> 2197 <div class="author"> 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 <em>Kamarthi, Harshavardhan</em>, 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 <a href="https://lingkai-kong.com" target="_blank" rel="noopener noreferrer">Kong, Lingkai</a>, 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 <a href="https://alrodri.engin.umich.edu/" target="_blank" rel="noopener noreferrer">Rodriguez, Alexander</a>, 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 <a href="https://chaozhang.org" target="_blank" rel="noopener noreferrer">Zhang, Chao</a>, 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 2287 2288 2289 2290 2291 2292 </div> 2293 2294 <div class="periodical"> 2295 2296 <em>KDD 2024 Workshop on Uncertainty Reasoning and Quantification in Decision Making</em> 2297 2298 2299 2024 2300 2301 </div> 2302 2303 2304 <div class="links"> 2305 2306 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 2307 2308 2309 2310 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 2311 2312 2313 2314 2315 <a href="https://arxiv.org/abs/2407.02641" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 2316 2317 2318 2319 2320 2321 <a href="https://github.com/AdityaLab/STOIC" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">Code</a> 2322 2323 2324 2325 2326 </div> 2327 2328 <!-- Hidden abstract block --> 2329 2330 <div class="abstract hidden"> 2331 <p>Multi-variate time series forecasting is an important problem with a wide range of applications. Recent works model the relations between time-series as graphs and have shown that propagating information over the relation graph can improve time series forecasting. However, in many cases, relational information is not available or is noisy and reliable. Moreover, most works ignore the underlying uncertainty of time-series both for structure learning and deriving the forecasts resulting in the structure not capturing the uncertainty resulting in forecast distributions with poor uncertainty estimates. We tackle this challenge and introduce STOIC, that leverages stochastic correlations between time-series to learn underlying structure between time-series and to provide well-calibrated and accurate forecasts. Over a wide-range of benchmark datasets STOIC provides around 16% more accurate and 14% better-calibrated forecasts. 2332 STOIC also shows better adaptation to noise in data during inference and captures important and useful relational information in various benchmarks.</p> 2333 </div> 2334 2335 2336 <!-- Hidden bibtex block --> 2337 2338 <div class="bibtex hidden"> 2339 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">kamarthi2024stoic</span><span class="p">,</span> 2340 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{stoic24}</span><span class="p">,</span> 2341 <span class="na">title</span> <span class="p">=</span> <span class="s">{Learning Graph Structures and Uncertainty for Accurate and Calibrated Time-series Forec
2341asting}</span><span class="p">,</span> 2342 <span class="na">author</span> <span class="p">=</span> <span class="s">{Kamarthi, Harshavardhan and Kong, Lingkai and Rodriguez, Alexander and Zhang, Chao and Prakash, B Aditya}</span><span class="p">,</span> 2343 <span class="na">journal</span> <span class="p">=</span> <span class="s">{KDD 2024 Workshop on Uncertainty Reasoning and Quantification in Decision Making}</span><span class="p">,</span> 2344 <span class="na">year</span> <span class="p">=</span> <span class="s">{2024}</span><span class="p">,</span> 2345 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 2346 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2407.02641}</span><span class="p">,</span> 2347 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 2348 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/AdityaLab/STOIC}</span> 2349<span class="p">}</span></code></pre></figure> 2350 </div> 2351 2352 </div> 2353</div> 2354</li> 2355<li> 2356<div class="row"> 2357 <div class="col-sm-2 abbr"> 2358 2359 <img class="img-fluid" src="/assets/pubimg/survey24.png" alt="survey24"> 2360 2361 </div> 2362 2363 <div id="rodriguez2024machine" class="col-sm-8"> 2364 2365 <div class="title">Machine learning for data-centric epidemic forecasting</div> 2366 <div class="author"> 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 RodrıÌguez, Alexander, 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 <em>Kamarthi, Harshavardhan</em>, 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 Agarwal, Pulak, 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 Ho, Javen, 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 Patel, Mira, 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 Sapre, Suchet, 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 2471 2472 2473 2474 2475 2476 </div> 2477 2478 <div class="periodical"> 2479 2480 <em>Nature Machine Intelligence</em> 2481 2482 2483 2024 2484 2485 </div> 2486 2487 2488 <div class="links"> 2489 2490 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 2491 2492 2493 2494 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 2495 2496 2497 2498 2499 <a href="https://www.nature.com/articles/s42256-024-00895-7" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 2500 2501 2502 2503 2504 2505 2506 2507 2508 </div> 2509 2510 <!-- Hidden abstract block --> 2511 2512 <div class="abstract hidden"> 2513 <p>The COVID-19 pandemic emphasized the importance of epidemic forecasting for decision makers in multiple domains, ranging from public health to the economy. Forecasting epidemic progression is a non-trivial task due to multiple confounding factors, such as human behaviour, pathogen dynamics and environmental conditions. However, the surge in research interest and initiatives from public health and funding agencies has fuelled the availability of new data sources that capture previously unobservable aspects of disease spread, paving the way for a spate of âdata-centredâ computational solutions that show promise for enhancing our forecasting capabilities. Here we discuss various methodological and practical advances and introduce a conceptual framework to navigate through them. First we list relevant datasets, such as symptomatic online surveys, retail and commerce, mobility and genomics data. Next we consider methods, focusing on re
2513cent data-driven statistical and deep learning-based methods, as well as hybrid models that combine domain knowledge of mechanistic models with the flexibility of statistical approaches. We also discuss experiences and challenges that arise in the real-world deployment of these forecasting systems, including decision-making informed by forecasts. Finally, we highlight some challenges and open problems found across the forecasting pipeline to enable robust future pandemic preparedness.</p> 2514 </div> 2515 2516 2517 <!-- Hidden bibtex block --> 2518 2519 <div class="bibtex hidden"> 2520 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">rodriguez2024machine</span><span class="p">,</span> 2521 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{survey24}</span><span class="p">,</span> 2522 <span class="na">title</span> <span class="p">=</span> <span class="s">{Machine learning for data-centric epidemic forecasting}</span><span class="p">,</span> 2523 <span class="na">author</span> <span class="p">=</span> <span class="s">{Rodr{\'\i}guez, Alexander and Kamarthi, Harshavardhan and Agarwal, Pulak and Ho, Javen and Patel, Mira and Sapre, Suchet and Prakash, B Aditya}</span><span class="p">,</span> 2524 <span class="na">journal</span> <span class="p">=</span> <span class="s">{Nature Machine Intelligence}</span><span class="p">,</span> 2525 <span class="na">volume</span> <span class="p">=</span> <span class="s">{6}</span><span class="p">,</span> 2526 <span class="na">number</span> <span class="p">=</span> <span class="s">{10}</span><span class="p">,</span> 2527 <span class="na">pages</span> <span class="p">=</span> <span class="s">{1122--1131}</span><span class="p">,</span> 2528 <span class="na">year</span> <span class="p">=</span> <span class="s">{2024}</span><span class="p">,</span> 2529 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 2530 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://www.nature.com/articles/s42256-024-00895-7}</span><span class="p">,</span> 2531 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span> 2532<span class="p">}</span></code></pre></figure> 2533 </div> 2534 2535 </div> 2536</div> 2537</li> 2538<li> 2539<div class="row"> 2540 <div class="col-sm-2 abbr"> 2541 2542 <img class="img-fluid" src="/assets/pubimg/pems23.png" alt="pems23"> 2543 2544 </div> 2545 2546 <div id="kamarthi2023pems" class="col-sm-8"> 2547 2548 <div class="title">PEMS: Pre-trained Epidmic Time-series Models</div> 2549 <div class="author"> 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 <em>Kamarthi, Harshavardhan</em>, 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 2579 2580 2581 2582 2583 2584 </div> 2585 2586 <div class="periodical"> 2587 2588 <em>arXiv preprint arXiv:2311.07841</em> 2589 2590 2591 2023 2592 2593 </div> 2594 2595 2596 <div class="links"> 2597 2598 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 2599 2600 2601 2602 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 2603 2604 2605 2606 2607 <a href="https://arxiv.org/abs/2311.07841" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 2608 2609 2610 2611 2612 2613 2614 2615 2616 </div> 2617 2618 <!-- Hidden abstract block --> 2619 2620 <div class="abstract hidden"> 2621 <p>Providing accurate and reliable predictions about the future of an epidemic is an important problem for enabling informed public health decisions. Recent works have shown that leveraging data-driven solutions that utilize advances in deep learning methods to learn from past data of an epidemic often outperform traditional mechanistic models. However, in many cases, the past data is sparse and may not sufficiently capture the underlying dynamics. While there exists a large amount of data from past epidemics, leveraging prior knowledge from time-series data of other diseases is a non-trivial challenge. Motivated by the success of pre-trained models in language and vision tasks, we tackle the problem of pre-training epidemic time-series models to learn from multiple datasets from different diseases and epidemics. We introduce Pre-trained Epidemic Time-Series Models (PEMS) that learn from diverse time-series datasets of a variety of diseases by formulating pre-training as a set of self-supervised learning (SSL) tasks. We tackle various important challenges specific to pre-training for epidemic time-series such as dealing with heterogeneous dynamics and efficie
2621ntly capturing useful patterns from multiple epidemic datasets by carefully designing the SSL tasks to learn important priors about the epidemic dynamics that can be leveraged for fine-tuning to multiple downstream tasks. The resultant PEM outperforms previous state-of-the-art methods in various downstream time-series tasks across datasets of varying seasonal patterns, geography, and mechanism of contagion including the novel Covid-19 pandemic unseen in pre-trained data with better efficiency using smaller fraction of datasets.</p> 2622 </div> 2623 2624 2625 <!-- Hidden bibtex block --> 2626 2627 <div class="bibtex hidden"> 2628 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">kamarthi2023pems</span><span class="p">,</span> 2629 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{pems23}</span><span class="p">,</span> 2630 <span class="na">title</span> <span class="p">=</span> <span class="s">{PEMS: Pre-trained Epidmic Time-series Models}</span><span class="p">,</span> 2631 <span class="na">author</span> <span class="p">=</span> <span class="s">{Kamarthi, Harshavardhan and Prakash, B Aditya}</span><span class="p">,</span> 2632 <span class="na">journal</span> <span class="p">=</span> <span class="s">{arXiv preprint arXiv:2311.07841}</span><span class="p">,</span> 2633 <span class="na">year</span> <span class="p">=</span> <span class="s">{2023}</span><span class="p">,</span> 2634 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 2635 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2311.07841}</span><span class="p">,</span> 2636 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span> 2637<span class="p">}</span></code></pre></figure> 2638 </div> 2639 2640 </div> 2641</div> 2642</li> 2643<li> 2644<div class="row"> 2645 <div class="col-sm-2 abbr"> 2646 2647 <img class="img-fluid" src="/assets/pubimg/profhit23.png" alt="profhit23"> 2648 2649 </div> 2650 2651 <div id="kamarthi2023profhit" class="col-sm-8"> 2652 2653 <div class="title">PROFHIT: Probabilistic Robust Forecasting for Hierarchical Time-series</div> 2654 <div class="author"> 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 <em>Kamarthi, Harshavardhan</em>, 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 <a href="https://lingkai-kong.com" target="_blank" rel="noopener noreferrer">Kong, Lingkai</a>, 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 RodrıÌguez, Alexander, 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 <a href="https://chaozhang.org" target="_blank" rel="noopener noreferrer">Zhang, Chao</a>, 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 2739 2740 2741 2742 2743 2744 </div> 2745 2746 <div class="periodical"> 2747 2748 <em>KDD</em> 2749 2750 2751 2023 2752 2753 </div> 2754 2755 2756 <div class="links"> 2757 2758 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 2759 2760 2761 2762 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 2763 2764 2765 2766 2767 <a href="https://arxiv.org/abs/2206.07940" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 2768 2769 2770 2771 2772 2773 <a href="https://github.com/AdityaLab/CAMul" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">
2773Code</a> 2774 2775 2776 2777 2778 </div> 2779 2780 <!-- Hidden abstract block --> 2781 2782 <div class="abstract hidden"> 2783 <p>Many multivariate time-series that have underlying hierarchical relations. Probabilistic hierarchical time-series forecasting is an important variant of time-series forecasting, where the goal is to model and forecast multivariate time-series. We propose PROFHIT, a state-of-art probabilistic model that provides accurate probabilitic forecasts that adhere to hierarchical relations at distributional level and adapts to datasets with different levels of adherance to the hierarchy.</p> 2784 </div> 2785 2786 2787 <!-- Hidden bibtex block --> 2788 2789 <div class="bibtex hidden"> 2790 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">kamarthi2023profhit</span><span class="p">,</span> 2791 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{profhit23}</span><span class="p">,</span> 2792 <span class="na">title</span> <span class="p">=</span> <span class="s">{PROFHIT: Probabilistic Robust Forecasting for Hierarchical Time-series}</span><span class="p">,</span> 2793 <span class="na">author</span> <span class="p">=</span> <span class="s">{Kamarthi, Harshavardhan and Kong, Lingkai and Rodr{\'\i}guez, Alexander and Zhang, Chao and Prakash, B Aditya}</span><span class="p">,</span> 2794 <span class="na">journal</span> <span class="p">=</span> <span class="s">{KDD}</span><span class="p">,</span> 2795 <span class="na">year</span> <span class="p">=</span> <span class="s">{2023}</span><span class="p">,</span> 2796 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 2797 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2206.07940}</span><span class="p">,</span> 2798 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 2799 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/AdityaLab/CAMul}</span> 2800<span class="p">}</span></code></pre></figure> 2801 </div> 2802 2803 </div> 2804</div> 2805</li> 2806<li> 2807<div class="row"> 2808 <div class="col-sm-2 abbr"> 2809 2810 <img class="img-fluid" src="/assets/pubimg/camul21.png" alt="camul21"> 2811 2812 </div> 2813 2814 <div id="kamarthi2021camul" class="col-sm-8"> 2815 2816 <div class="title">CAMul: Calibrated and Accurate Multi-view Time-Series Forecasting</div> 2817 <div class="author"> 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 <em>Kamarthi, Harshavardhan</em>, 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 <a href="https://lingkai-kong.com" target="_blank" rel="noopener noreferrer">Kong, Lingkai</a>, 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 RodrıÌguez, Alexander, 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 <a href="https://chaozhang.org" target="_blank" rel="noopener noreferrer">Zhang, Chao</a>, 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 2902 2903 2904 2905 2906 2907 </div> 2908 2909 <div class="periodical"> 2910 2911 <em>ACM Web Conference (WWW)</em> 2912 2913 2914 2022 2915 2916 </div> 2917 2918 2919 <div class="links"> 2920 2921 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 2922 2923 2924 2925 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 2926 2927 2928 2929 2930 <a href="https://arxiv.org/abs/2109.07438" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 2931 2932 2933 2934 2935 2936 <a href="https://github.com/AdityaLab/CAMul" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">
2936Code</a> 2937 2938 2939 2940 2941 </div> 2942 2943 <!-- Hidden abstract block --> 2944 2945 <div class="abstract hidden"> 2946 <p>Probabilistic time-series forecasting enables reliable decision making across many domains. Most forecasting problems have diverse sources of data containing multiple modalities and structures. Leveraging information as well as uncertainty from these data sources for well-calibrated and accurate forecasts is an important challenging problem. Most previous work on multi-modal learning and forecasting simply aggregate intermediate representations from each data view by simple methods of summation or concatenation and do not explicitly model uncertainty for each data-view. We propose a general probabilistic multi-view forecasting framework CAMul, that can learn representations and uncertainty from diverse data sources. It integrates the knowledge and uncertainty from each data view in a dynamic context-specific manner assigning more importance to useful views to model a well-calibrated forecast distribution. We use CAMul for multiple domains with varied sources and modalities and show that CAMul outperforms other state-of-art probabilistic forecasting models by over 25% in accuracy and calibration.</p> 2947 </div> 2948 2949 2950 <!-- Hidden bibtex block --> 2951 2952 <div class="bibtex hidden"> 2953 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">kamarthi2021camul</span><span class="p">,</span> 2954 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{camul21}</span><span class="p">,</span> 2955 <span class="na">title</span> <span class="p">=</span> <span class="s">{CAMul: Calibrated and Accurate Multi-view Time-Series Forecasting}</span><span class="p">,</span> 2956 <span class="na">author</span> <span class="p">=</span> <span class="s">{Kamarthi, Harshavardhan and Kong, Lingkai and Rodr{\'\i}guez, Alexander and Zhang, Chao and Prakash, B Aditya}</span><span class="p">,</span> 2957 <span class="na">journal</span> <span class="p">=</span> <span class="s">{ACM Web Conference (WWW)}</span><span class="p">,</span> 2958 <span class="na">year</span> <span class="p">=</span> <span class="s">{2022}</span><span class="p">,</span> 2959 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 2960 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2109.07438}</span><span class="p">,</span> 2961 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 2962 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/AdityaLab/CAMul}</span> 2963<span class="p">}</span></code></pre></figure> 2964 </div> 2965 2966 </div> 2967</div> 2968</li> 2969<li> 2970<div class="row"> 2971 <div class="col-sm-2 abbr"> 2972 2973 <img class="img-fluid" src="/assets/pubimg/back2future21.png" alt="back2future21"> 2974 2975 </div> 2976 2977 <div id="kamarthi2021back2future" class="col-sm-8"> 2978 2979 <div class="title">Back2Future: Leveraging Backfill Dynamics for Improving Real-time Predictions in Future</div> 2980 <div class="author"> 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 <em>Kamarthi, Harshavardhan</em>, 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 RodrıÌguez, Alexander, 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 3025 3026 3027 3028 3029 3030 </div> 3031 3032 <div class="periodical"> 3033 3034 <em>ICLR</em> 3035 3036 3037 2022 3038 3039 </div> 3040 3041 3042 <div class="links"> 3043 3044 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 3045 3046 3047 3048 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 3049 3050 3051 3052 3053 <a href="https://arxiv.org/abs/2106.04420" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 3054 3055 3056 3057 3058 3059 <a href="https://github.com/AdityaLab/Back2Future" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">
3059Code</a> 3060 3061 3062 3063 3064 </div> 3065 3066 <!-- Hidden abstract block --> 3067 3068 <div class="abstract hidden"> 3069 <p>In real-time forecasting in public health, data collection is a non-trivial and demanding task. Often after initially released, it undergoes several revisions later (maybe due to human or technical constraints) - as a result, it may take weeks until the data reaches to a stable value. This so-called âbackfillâ phenomenon and its effect on model performance has been barely studied in the prior literature. In this paper, we introduce the multi-variate backfill problem using COVID-19 as the motivating example. We construct a detailed dataset composed of relevant signals over the past year of the pandemic. We then systematically characterize several patterns in backfill dynamics and leverage our observations for formulating a novel problem and neural framework Back2Future that aims to refines a given modelâs predictions in real-time. Our extensive experiments demonstrate that our method refines the performance of top models for COVID-19 forecasting, in contrast to non-trivial baselines, yielding 18% improvement over baselines, enabling us obtain a new SOTA performance. In addition, we show that our model improves model evaluation too; hence policy-makers can better understand the true accuracy of forecasting models in real-time.</p> 3070 </div> 3071 3072 3073 <!-- Hidden bibtex block --> 3074 3075 <div class="bibtex hidden"> 3076 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">kamarthi2021back2future</span><span class="p">,</span> 3077 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{back2future21}</span><span class="p">,</span> 3078 <span class="na">title</span> <span class="p">=</span> <span class="s">{Back2Future: Leveraging Backfill Dynamics for Improving Real-time Predictions in Future}</span><span class="p">,</span> 3079 <span class="na">author</span> <span class="p">=</span> <span class="s">{Kamarthi, Harshavardhan and Rodr{\'\i}guez, Alexander and Prakash, B Aditya}</span><span class="p">,</span> 3080 <span class="na">journal</span> <span class="p">=</span> <span class="s">{ICLR}</span><span class="p">,</span> 3081 <span class="na">year</span> <span class="p">=</span> <span class="s">{2022}</span><span class="p">,</span> 3082 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 3083 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2106.04420}</span><span class="p">,</span> 3084 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 3085 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/AdityaLab/Back2Future}</span> 3086<span class="p">}</span></code></pre></figure> 3087 </div> 3088 3089 </div> 3090</div> 3091</li> 3092<li> 3093<div class="row"> 3094 <div class="col-sm-2 abbr"> 3095 3096 <img class="img-fluid" src="/assets/pubimg/epifnp21.png" alt="epifnp21"> 3097 3098 </div> 3099 3100 <div id="kamarthi2021doubt" class="col-sm-8"> 3101 3102 <div class="title">When in Doubt: Neural Non-Parametric Uncertainty Quantification for Epidemic Forecasting</div> 3103 <div class="author"> 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 <em>Kamarthi, Harshavardhan</em>, 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 <a href="https://lingkai-kong.com" target="_blank" rel="noopener noreferrer">Kong, Lingkai</a>, 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 RodrıÌguez, Alexander, 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 <a href="https://chaozhang.org" target="_blank" rel="noopener noreferrer">Zhang, Chao</a>, 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 and <a href="https://cc.gatech.edu/~badityap" target="_blank" rel="noopener noreferrer">Prakash, B Aditya</a> 3188 3189 3190 3191 3192 3193 </div> 3194 3195 <div class="periodical"> 3196 3197 <em>Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS)</em> 3198 3199 3200 2021 3201 3202 </div> 3203 3204 3205 <div class="links"> 3206 3207 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 3208 3209 3210 3211 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 3212 3213 3214 3215 3216 <a href="https://arxiv.org/abs/2106.03904" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 3217 3218 3219 3220 3221 3222 <a href="https://github.com/AdityaLab/EpiFNP" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">
3222Code</a> 3223 3224 3225 3226 3227 </div> 3228 3229 <!-- Hidden abstract block --> 3230 3231 <div class="abstract hidden"> 3232 <p>Accurate and trustworthy epidemic forecasting is an important problem that has impact on public health planning and disease mitigation. Most existing epidemic forecasting models disregard uncertainty quantification, resulting in mis-calibrated predictions. Recent works in deep neural models for uncertainty-aware time-series forecasting also have several limitations; e.g. it is difficult to specify meaningful priors in Bayesian NNs, while methods like deep ensembling are computationally expensive in practice. In this paper, we fill this important gap. We model the forecasting task as a probabilistic generative process and propose a functional neural process model called EPIFNP, which directly models the probability density of the forecast value. EPIFNP leverages a dynamic stochastic correlation graph to model the correlations between sequences in a non-parametric way, and designs different stochastic latent variables to capture functional uncertainty from different perspectives. Our extensive experiments in a real-time flu forecasting setting show that EPIFNP significantly outperforms previous state-of-the-art models in both accuracy and calibration metrics, up to 2.5x in accuracy and 2.4x in calibration. Additionally, due to properties of its generative process,EPIFNP learns the relations between the current season and similar patterns of historical seasons,enabling interpretable forecasts. Beyond epidemic forecasting, the EPIFNP can be of independent interest for advancing principled uncertainty quantification in deep sequential models for predictive analytics.</p> 3233 </div> 3234 3235 3236 <!-- Hidden bibtex block --> 3237 3238 <div class="bibtex hidden"> 3239 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">kamarthi2021doubt</span><span class="p">,</span> 3240 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{epifnp21}</span><span class="p">,</span> 3241 <span class="na">title</span> <span class="p">=</span> <span class="s">{When in Doubt: Neural Non-Parametric Uncertainty Quantification for Epidemic Forecasting}</span><span class="p">,</span> 3242 <span class="na">author</span> <span class="p">=</span> <span class="s">{Kamarthi, Harshavardhan and Kong, Lingkai and Rodr{\'\i}guez, Alexander and Zhang, Chao and Prakash, B Aditya}</span><span class="p">,</span> 3243 <span class="na">journal</span> <span class="p">=</span> <span class="s">{Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS)}</span><span class="p">,</span> 3244 <span class="na">year</span> <span class="p">=</span> <span class="s">{2021}</span><span class="p">,</span> 3245 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 3246 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2106.03904}</span><span class="p">,</span> 3247 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 3248 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/AdityaLab/EpiFNP}</span> 3249<span class="p">}</span></code></pre></figure> 3250 </div> 3251 3252 </div> 3253</div> 3254</li> 3255<li> 3256<div class="row"> 3257 <div class="col-sm-2 abbr"> 3258 3259 <img class="img-fluid" src="/assets/pubimg/selective20.png" alt="selective20"> 3260 3261 </div> 3262 3263 <div id="nishtala2021selective" class="col-sm-8"> 3264 3265 <div class="title">Selective Intervention Planning using Restless Multi-Armed Bandits to Improve Maternal and Child Health Outcomes</div> 3266 <div class="author"> 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 Nishtala, Siddharth, 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 Madaan, Lovish, 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 Mate, Aditya, 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 <em>Kamarthi, Harshavardhan</em>, 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 Grama, Anirudh, 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 Thakkar, Divy, 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 Narayanan, Dhyanesh, 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 Chaudhary, Suresh, 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 Madhiwalla, Neha, 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 Padmanabhan, Ramesh, 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 and others, 3426 3427 3428 3429 3430 3431 </div> 3432 3433 <div class="periodical"> 3434 3435 <em>arXiv preprint arXiv:2103.09052</em> 3436 3437 3438 2021 3439 3440 </div> 3441 3442 3443 <div class="links"> 3444 3445 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 3446 3447 3448 3449 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 3450 3451 3452 3453 3454 <a href="https://arxiv.org/abs/2103.09052" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 3455 3456 3457 3458 3459 3460 3461 3462 3463 </div> 3464 3465 <!-- Hidden abstract block --> 3466 3467 <div class="abstract hidden"> 3468 <p>India has a maternal mortality ratio of 113 and child mortality ratio of 2830 per 100,000 live births. Lack of access to preventive care information is a major contributing factor for these deaths, especially in low resource households. We partner with ARMMAN, a non-profit based in India employing a call-based information program to disseminate health-related information to pregnant women and women with recent child deliveries. We analyze call records of over 300,000 women registered in the program created by ARMMAN and try to identify women who might not engage with these call programs that are proven to result in positive health outcomes. We built machine learning based models to predict the long term engagement pattern from call logs and beneficiariesâ demographic information, and discuss the applicability of this method in the real world through a pilot validation. Through a randomized controlled trial, we show that using our modelâs predictions to make interventions boosts engagement metrics by 61.37%. We then formulate the intervention planning problem as restless multi-armed bandits (RMABs), and present preliminary results using this approach.</p> 3469 </div> 3470 3471 3472 <!-- Hidden bibtex block --> 3473 3474 <div class="bibtex hidden"> 3475 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">nishtala2021selective</span><span class="p">,</span> 3476 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{selective20}</span><span class="p">,</span> 3477 <span class="na">title</span> <span class="p">=</span> <span class="s">{Selective Intervention Planning using Restless Multi-Armed Bandits to Improve Maternal and Child Health Outcomes}</span><span class="p">,</span> 3478 <span class="na">author</span> <span class="p">=</span> <span class="s">{Nishtala, Siddharth and Madaan, Lovish and Mate, Aditya and Kamarthi, Harshavardhan and Grama, Anirudh and Thakkar, Divy and Narayanan, Dhyanesh and Chaudhary, Suresh and Madhiwalla, Neha and Padmanabhan, Ramesh and others}</span><span class="p">,</span> 3479 <span class="na">
3479journal</span> <span class="p">=</span> <span class="s">{arXiv preprint arXiv:2103.09052}</span><span class="p">,</span> 3480 <span class="na">year</span> <span class="p">=</span> <span class="s">{2021}</span><span class="p">,</span> 3481 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2103.09052}</span><span class="p">,</span> 3482 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 3483 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span> 3484<span class="p">}</span></code></pre></figure> 3485 </div> 3486 3487 </div> 3488</div> 3489</li> 3490<li> 3491<div class="row"> 3492 <div class="col-sm-2 abbr"> 3493 3494 <img class="img-fluid" src="/assets/pubimg/patrol20.png" alt="patrol20"> 3495 3496 </div> 3497 3498 <div id="venugopal2020reinforcement" class="col-sm-8"> 3499 3500 <div class="title">Reinforcement Learning for Unified Allocation and Patrolling in Signaling Games with Uncertainty</div> 3501 <div class="author"> 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 Venugopal, Aravind, 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 <a href="https://sites.google.com/view/elizabethbondi" target="_blank" rel="noopener noreferrer">Bondi, Elizabeth</a>, 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 <em>Kamarthi, Harshavardhan</em>, 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 Dholakia, Keval, 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 <a href="https://www.cse.iitm.ac.in/~ravi/" target="_blank" rel="noopener noreferrer">Ravindran, Balaraman</a>, 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 and <a href="https://teamcore.seas.harvard.edu/tambe" target="_blank" rel="noopener noreferrer">Tambe, Milind</a> 3601 3602 3603 3604 3605 3606 </div> 3607 3608 <div class="periodical"> 3609 3610 <em>20th International Conference on Autonomous Agents and Multiagent Systems (AAMAS)</em> 3611 3612 3613 2020 3614 3615 </div> 3616 3617 3618 <div class="links"> 3619 3620 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 3621 3622 3623 3624 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 3625 3626 3627 3628 3629 <a href="https://arxiv.org/abs/2012.10389" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 3630 3631 3632 3633 3634 3635 3636 3637 3638 </div> 3639 3640 <!-- Hidden abstract block --> 3641 3642 <div class="abstract hidden"> 3643 <p>Green Security Games (GSGs) have been successfully used in the protection of valuable resources such as fisheries, forests and wildlife. While real-world deployment involves both resource allocation and subsequent coordinated patrolling with communication and real-time, uncertain information, previous game models do not fully address both of these stages simultaneously. Furthermore, adopting existing solution strategies is difficult since they do not scale well for larger, more complex variants of the game models. 3644 We therefore first propose a novel GSG model that combines defender allocation, patrolling, real-time drone notification to human patrollers, and drones sending warning signals to attackers. The model further incorporates uncertainty for real-time decision-making within a team of drones and human patrollers. Second, we present CombSGPO, a novel and scalable algorithm based on reinforcement learning, to compute a defender strategy for this game model. CombSGPO performs policy search over a multi-dimensional, discrete action space to compute an allocation strategy that is best suited to a best-response patrolling strategy for the defender, learnt by training a multi-agent Deep Q-Network. We show via experiments that CombSGPO converges to better strategies and is more scalable than comparable approaches. Third, we provide a detailed analysis of the coordination and signaling behavior learnt by CombSGPO, showing group formation between defender resources and patrolling formations based on signaling and notifications between resources. Importantly, we find that strategic signaling emerges in the final learnt strategy. Finally, we perform experiments to evaluate these strategies under different levels of uncertainty.</p> 3645 </div> 3646 3647 3648 <!-- Hidden bibtex block --> 3649 3650 <div class="bibtex hidden"> 3651 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">venugopal2020reinforcement</span><span class="p">,</span> 3652 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{patrol20}</span><span class="p">,</span> 3653 <span class="na">title</span> <span class="p">=</span> <span class="s">{Reinforcement Learning for Unified Allocation and Patrolling in Signaling Games with Uncertainty}</span><span class="p">,</span> 3654 <span class="na">author</span> <span class="p">=</span> <span class="s">{Venugopal, Aravind and Bondi, Elizabeth and Kamarthi, Harshavardhan and Dholakia, Keval and Ravindran, Balaraman and Tambe, Milind}</span><span class="p">,</span> 3655 <span class="na">
3655journal</span> <span class="p">=</span> <span class="s">{20th International Conference on Autonomous Agents and Multiagent Systems (AAMAS)}</span><span class="p">,</span> 3656 <span class="na">year</span> <span class="p">=</span> <span class="s">{2020}</span><span class="p">,</span> 3657 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2012.10389}</span><span class="p">,</span> 3658 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 3659 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span> 3660<span class="p">}</span></code></pre></figure> 3661 </div> 3662 3663 </div> 3664</div> 3665</li> 3666<li> 3667<div class="row"> 3668 <div class="col-sm-2 abbr"> 3669 3670 <img class="img-fluid" src="/assets/pubimg/missedcall20.png" alt="missedcall20"> 3671 3672 </div> 3673 3674 <div id="nishtala2020missed" class="col-sm-8"> 3675 3676 <div class="title">Missed calls, Automated Calls and Health Support: Using AI to improve maternal health outcomes by increasing program engagement</div> 3677 <div class="author"> 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 Nishtala, Siddharth, 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 <em>Kamarthi, Harshavardhan</em>, 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 Thakkar, Divy, 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 Narayanan, Dhyanesh, 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 Grama, Anirudh, 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 Hegde, Aparna, 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 Padmanabhan, Ramesh, 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 Madhiwalla, Neha, 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 Chaudhary, Suresh, 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 <a href="https://www.cse.iitm.ac.in/~ravi/" target="_blank" rel="noopener noreferrer">Ravindran, Balaraman</a>, 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 and others, 3842 3843 3844 3845 3846 3847 </div> 3848 3849 <div class="periodical"> 3850 3851 <em>Harvard CRCS Workshop on AI for Social Good</em> 3852 3853 3854 2020 3855 3856 </div> 3857 3858 3859 <div class="links"> 3860 3861 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 3862 3863 3864 3865 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 3866 3867 3868 3869 3870 <a href="https://arxiv.org/abs/2006.07590" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 3871 3872 3873 3874 3875 3876 3877 3878 3879 </div> 3880 3881 <!-- Hidden abstract block --> 3882 3883 <div class="abstract hidden"> 3884 <p>India accounts for 11% of maternal deaths globally where a woman dies in childbirth every fifteen minutes. Lack of access to preventive care information is a significant problem contributing to high maternal morbidity and mortality numbers, especially in low-income households. We work with ARMMAN, a non-profit based in India, to further the use of call-based information programs by early-on identifying women who might not engage on these programs that are proven to affect health parameters positively.We analyzed anonymized call-records of over 300,000 women registered in an awareness program created by ARMMAN that uses cellphone calls to regularly disseminate health related information. We built robust deep learning based models to predict short term and long term dropout risk from call logs and beneficiariesâ demographic information. Our model performs 13% better than competitive baselines for short-term forecasting and 7% better for long term forecasting. We also discuss the applicability of this method in the real world through a pilot validation that uses our method to perform targeted interventions.</p> 3885 </div> 3886 3887 3888 <!-- Hidden bibtex block --> 3889 3890 <div class="bibtex hidden"> 3891 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">nishtala2020missed</span><span class="p">,</span> 3892 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{missedcall20}</span><span class="p">,</span> 3893 <span class="na">title</span> <span class="p">=</span> <span class="s">{Missed calls, Automated Calls and Health Support: Using AI to improve maternal health outcomes by increasing program engagement}</span><span class="p">,</span> 3894 <span class="na">author</span> <span class="p">=</span> <span class="s">{Nishtala, Siddharth and Kamarthi, Harshavardhan and Thakkar, Divy and Narayanan, Dhyanesh and Grama, Anirudh and Hegde, Aparna and Padmanabhan, Ramesh and Madhiwalla, Neha and Chaudhary, Suresh and Ravindran, Balaraman and others}</span><span class="p">,</span> 3895 <span class="na">
3895journal</span> <span class="p">=</span> <span class="s">{Harvard CRCS Workshop on AI for Social Good}</span><span class="p">,</span> 3896 <span class="na">year</span> <span class="p">=</span> <span class="s">{2020}</span><span class="p">,</span> 3897 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 3898 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 3899 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2006.07590}</span> 3900<span class="p">}</span></code></pre></figure> 3901 </div> 3902 3903 </div> 3904</div> 3905</li> 3906<li> 3907<div class="row"> 3908 <div class="col-sm-2 abbr"> 3909 3910 <img class="img-fluid" src="/assets/pubimg/influence19.png" alt="influence19"> 3911 3912 </div> 3913 3914 <div id="kamarthi2019influence" class="col-sm-8"> 3915 3916 <div class="title">Influence maximization in unknown social networks: Learning Policies for Effective Graph Sampling</div> 3917 <div class="author"> 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 <em>Kamarthi, Harshavardhan</em>, 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 Vijayan, Priyesh, 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 Wilder, Bryan, 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 <a href="https://www.cse.iitm.ac.in/~ravi/" target="_blank" rel="noopener noreferrer">Ravindran, Balaraman</a>, 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 and <a href="https://teamcore.seas.harvard.edu/tambe" target="_blank" rel="noopener noreferrer">Tambe, Milind</a> 3997 3998 3999 4000 4001 4002 </div> 4003 4004 <div class="periodical"> 4005 4006 <em>19th International Conference on Autonomous Agents and Multiagent Systems (AAMAS), <b>Nominated for Best Paper Award<\b> </b></em> 4007 4008 4009 2019 4010 4011 </div> 4012 4013 4014 <div class="links"> 4015 4016 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 4017 4018 4019 4020 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 4021 4022 4023 4024 4025 <a href="https://arxiv.org/abs/1907.11625" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 4026 4027 4028 4029 4030 4031 <a href="https://github.com/kage08/graph_sample_rl" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">Code</a> 4032 4033 4034 4035 4036 </div> 4037 4038 <!-- Hidden abstract block --> 4039 4040 <div class="abstract hidden"> 4041 <p>A serious challenge when finding influential actors in real-world social networks is the lack of knowledge about the structure of the underlying network. Current state-of-the-art methods rely on hand-crafted sampling algorithms; these methods sample nodes and their neighbours in a carefully constructed order and choose opinion leaders from this discovered network to maximize influence spread in the (unknown) complete network. In this work, we propose a reinforcement learning framework for network discovery that automatically learns useful node and graph representations that encode important structural properties of the network. At training time, the method identifies portions of the network such that the nodes selected from this sampled subgraph can effectively influence nodes in the complete network. The realization of such transferable network structure based adaptable policies is attributed to the meticulous design of the framework that encodes relevant node and graph signatures driven by an appropriate reward scheme. We experiment with real-world social networks from four different domains and show that the policies learned by our RL agent provide a 10-36% improvement over the current state-of-the-art method.</p> 4042 </div> 4043 4044 4045 <!-- Hidden bibtex block --> 4046 4047 <div class="bibtex hidden"> 4048 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">kamarthi2019influence</span><span class="p">,</span> 4049 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{influence19}</span><span class="p">,</span> 4050 <span class="na">title</span> <span class="p">=</span> <span class="s">{Influence maximization in unknown social networks: Learning Policies for Effective Graph Sampling}</span><span class="p">,</span> 4051 <span class="na">author</span> <span class="p">=</span> <span class="s">{Kamarthi, Harshavardhan and Vijayan, Priyesh and Wilder, Bryan and Ravindran, Balaraman and Tambe, Milind}</span><span class="p">,</span> 4052 <span class="na">
4052journal</span> <span class="p">=</span> <span class="s">{19th International Conference on Autonomous Agents and Multiagent Systems (AAMAS), <b>Nominated for Best Paper Award<\b> }</span><span class="p">,</span> 4053 <span class="na">note</span> <span class="p">=</span> <span class="s">{Nominated for best paper award}</span><span class="p">,</span> 4054 <span class="na">year</span> <span class="p">=</span> <span class="s">{2019}</span><span class="p">,</span> 4055 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 4056 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/1907.11625}</span><span class="p">,</span> 4057 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 4058 <span class="na">code</span> <span class="p">=</span> <span class="s">{https://github.com/kage08/graph_sample_rl}</span> 4059<span class="p">}</span></code></pre></figure> 4060 </div> 4061 4062 </div> 4063</div> 4064</li> 4065<li> 4066<div class="row"> 4067 <div class="col-sm-2 abbr"> 4068 4069 <img class="img-fluid" src="/assets/pubimg/integrating19.png" alt="integrating19"> 4070 4071 </div> 4072 4073 <div id="srinivasan2019integrating" class="col-sm-8"> 4074 4075 <div class="title">Integrating Lexical Knowledge in Word Embeddings using Sprinkling and Retrofitting</div> 4076 <div class="author"> 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 Srinivasan, Aakash, 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 <em>Kamarthi, Harshavardhan</em>, 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 Ganesan, Devi, 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 and Chakraborti, Sutanu 4131 4132 4133 4134 4135 4136 </div> 4137 4138 <div class="periodical"> 4139 4140 <em>International Conference on Natural Language Processing (ICNLP)</em> 4141 4142 4143 2019 4144 4145 </div> 4146 4147 4148 <div class="links"> 4149 4150 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 4151 4152 4153 4154 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 4155 4156 4157 4158 4159 <a href="https://arxiv.org/pdf/1912.06889" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="noopener noreferrer">PDF</a> 4160 4161 4162 4163 4164 4165 4166 4167 4168 </div> 4169 4170 <!-- Hidden abstract block --> 4171 4172 <div class="abstract hidden"> 4173 <p>Neural network based word embeddings, such as Word2Vec and GloVe, are purely data driven in that they capture the distributional information about words from the training corpus. Past works have attempted to improve these embeddings by incorporating semantic knowledge from lexical resources like WordNet. Some techniques like retrofitting modify word embeddings in the post-processing stage while some others use a joint learning approach by modifying the objective function of neural networks. In this paper, we discuss two novel approaches for incorporating semantic knowledge into word embeddings. In the first approach, we take advantage of Levy et alâs work which showed that using SVD based methods on co-occurrence matrix provide similar performance to neural network based embeddings. We propose a âsprinklingâ technique to add semantic relations to the co-occurrence matrix directly before factorization. In the second approach, WordNet similarity scores are used to improve the retrofitting method. We evaluate the proposed methods in both intrinsic and extrinsic tasks and observe significant improvements over the baselines in many of the datasets.</p> 4174 </div> 4175 4176 4177 <!-- Hidden bibtex block --> 4178 4179 <div class="bibtex hidden"> 4180 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">srinivasan2019integrating</span><span class="p">,</span> 4181 <span class="na">abbr</span> <span class="p">=</span> <span class="s">{integrating19}</span><span class="p">,</span> 4182 <span class="na">title</span> <span class="p">=</span> <span class="s">{Integrating Lexical Knowledge in Word Embeddings using Sprinkling and Retrofitting}</span><span class="p">,</span> 4183 <span class="na">author</span> <span class="p">=</span> <span class="s">{Srinivasan, Aakash and Kamarthi, Harshavardhan and Ganesan, Devi and Chakraborti, Sutanu}</span><span class="p">,</span> 4184 <span class="na">
4184journal</span> <span class="p">=</span> <span class="s">{International Conference on Natural Language Processing (ICNLP)}</span><span class="p">,</span> 4185 <span class="na">year</span> <span class="p">=</span> <span class="s">{2019}</span><span class="p">,</span> 4186 <span class="na">bibtex_show</span> <span class="p">=</span> <span class="s">{true}</span><span class="p">,</span> 4187 <span class="na">pdf</span> <span class="p">=</span> <span class="s">{https://arxiv.org/pdf/1912.06889}</span><span class="p">,</span> 4188 <span class="na">selected</span> <span class="p">=</span> <span class="s">{true}</span> 4189<span class="p">}</span></code></pre></figure> 4190 </div> 4191 4192 </div> 4193</div> 4194</li> 4195</ol> 4196</div> 4197 4198 4199 4200 <div class="social"> 4201 <div class="contact-icons"> 4202 <a href="mailto:%68%61%72%73%68%61%76%61%72%64%68%61%6E%38%36%34.%68%6B@%67%6D%61%69%6C.%63%6F%6D"><i class="fas fa-envelope"></i></a> 4203 4204<a href="https://scholar.google.com/citations?user=https://scholar.google.com/citations?user=LNXEjT8AAAAJ&hl=en" title="Google Scholar" target="_blank" rel="noopener noreferrer"><i class="ai ai-google-scholar"></i></a> 4205 4206 4207<a href="https://github.com/kage08" title="GitHub" target="_blank" rel="noopener noreferrer"><i class="fab fa-github"></i></a> 4208 4209<a href="https://twitter.com/harsha_64" title="Twitter" target="_blank" rel="noopener noreferrer"><i class="fab fa-twitter"></i></a> 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221<a href="https://kage08.github.io/feed.xml" title="RSS Feed"><i class="fas fa-rss-square"></i></a> 4222 4223 </div> 4224 <div class="contact-note">Best way to get in touch is via email! 4225</div> 4226 </div> 4227 4228 </article> 4229 4230</div> 4231 4232 </div> 4233 4234 <!-- Footer --> 4235 4236 4237<footer class="sticky-bottom mt-5"> 4238 <div class="container"> 4239 © Copyright 2026 Harshavardhan P Kamarthi. 4240 4241 4242 4243 Last updated: April 24, 2026. 4244 4245 </div> 4246</footer> 4247 4248 4249 4250 </body> 4251 4252 <!-- Bootsrap & MDB scripts -->
4253<script src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/2.4.4/umd/popper.min.js" integrity="sha512-eUQ9hGdLjBjY3F41CScH3UX+4JDSI9zXeroz7hJ+RteoCaY+GP/LDoM8AO+Pt+DRFw3nXqsjh9Zsts8hnYv8/A==" crossorigin="anonymous"></script>
vendor: 1 bytes, line 4253
4253
4254<script src="https://stackpath.bootstrapcdn.com/bootstrap/4.5.2/js/bootstrap.min.js" integrity="sha512-M5KW3ztuIICmVIhjSqXe01oV2bpe248gOxqmlcYrEzAvws7Pw3z6BK0iGbrwvdrUQUhi3eXgtxp5I8PDo9YfjQ==" crossorigin="anonymous"></script>
vendor: 1 bytes, line 4254
4254
4255<script src="https://cdnjs.cloudflare.com/ajax/libs/mdbootstrap/4.19.1/js/mdb.min.js" integrity="sha512-Mug9KHKmroQFMLm93zGrjhibM2z2Obg9l6qFG2qKjXEXkMp/VDkI4uju9m4QKPjWSwQ6O2qzZEnJDEeCw0Blcw==" crossorigin="anonymous"></script>
4255 4256 4257 4258<!-- Mansory & imagesLoaded -->
4259<script defer src="https://unpkg.com/masonry-layout@4/dist/masonry.pkgd.min.js"></script>
vendor: 1 bytes, line 4259
4259
4260<script defer src="https://unpkg.com/imagesloaded@4/imagesloaded.pkgd.min.js"></script>
vendor: 1 bytes, line 4260
4260
4261<script defer src="/assets/js/mansory.js" type="text/javascript"></script>
4261 4262 4263 4264 4265 4266 4267<!-- Medium Zoom JS -->
4268<script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/medium-zoom.min.js" integrity="sha256-EdPgYcPk/IIrw7FYeuJQexva49pVRZNmt3LculEr7zM=" crossorigin="anonymous"></script>
vendor: 1 bytes, line 4268
4268
4269<script src="/assets/js/zoom.js"></script>
4269 4270 4271 4272<!-- Load Common JS -->
4273<script src="/assets/js/common.js"></script>
4273 4274 4275 4276</html>
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.