Implemented as a Multi-Layer Perceptron to classify handwritten Digits (0-9)
Model Architecture and Results
The model comprises a flattening layer and three linear layers ((256, 64) hidden dimensions) with relus to approximate non-linearity. It achieves 95.6% accuracy after 15 training epochs and batch size = 64. Taining and Test MNIST datasets are loaded with PyTorch dataloaders.