<MNIST_Demo/>

Draw a digit (0-9) and watch the neural network predict in real-time

Prediction: _

0
0.0%
1
0.0%
2
0.0%
3
0.0%
4
0.0%
5
0.0%
6
0.0%
7
0.0%
8
0.0%
9
0.0%

Powered by a neural network library I built from scratch in C++, running on the CPU. No TensorFlow, no PyTorch. Just raw matrix math, backpropagation, and optimization implemented by hand.

Architecture: 784-256-128-10 fully-connected network. ReLU activations, softmax output, cross-entropy loss.

Training: Adam optimizer (lr=0.001, β₁=0.9, β₂=0.999). Batch size 32. 96%+ accuracy on 10k test set.

Implementation: C++23. Multi-threaded with adaptive parallelism. ~3k samples/sec throughput.