fashion

The "nn-fashion" project on Kaggle builds and compares a deep neural network and a convolutional neural network (convNet) to classify Fashion MNIST images (70,000 grayscale 28x28 pixel images across 10 clothing categories). uses fully connected layers (784-512-256-10) with ReLU and dropout (0.2), while convNet applies two convolutional layers (1-16-32 channels) with max pooling, followed by fully connected layers (7x7x32-512-256-10) with dropout. Training runs for 50 epochs using PyTorch, Adam optimizer, and CrossEntropyLoss, with GPU support if available. Data is preprocessed with torchvision, and losses are tracked and visualized with matplotlib. The best model is saved based on validation loss, offering a practical comparison of network types for image classification.