1const e=`The Digit Recognizer is an end-to-end machine learning project: a convolutional neural network trained to identify handwritten digits, paired with a web app where you can draw a digit and watch the model's confidence scores update in real time. 2 3The model is an 8-layer CNN trained on the MNIST dataset (60,000 training images and 10,000 test images), achieving 99.3% accuracy. I used Matplotlib to visualize predictions during development, which made it much easier to understand where the model was struggling. 4 5## The web layer 6 7I wanted the model to feel interactive, not just a static demo. The frontend is built in React with a canvas element for drawing. As you draw, the predictions update live with a bar chart showing confidence for each digit class. It made the model's internals feel transparent. 8 9## What I explored 10 11I experimented with network depth, pooling strategies, and dropout rates to understand their effects on convergence and generalization. The jump from 97% to 99%+ accuracy came mostly from adding more convolutional layers and tuning the learning rate, which was simpler than I expected. 12 13The project gave me a solid foundation in the full ML workflow: data preprocessing, model design, training, evaluation, and deployment. 14`;export{e as default};
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.