About

An end-to-end implementation of a multi-layer perceptron (MLP) for MNIST digit classification, utilizing Kaiming weight initialization, ReLU activations, and Adam optimization—engineered with GPU acceleration support via PyTorch for efficient training and inference.