A multicategory classification model assigns each input to one of more than two discrete classes. Fashion MNIST provides a concrete case: the model distinguishes 10 clothing categories rather than choosing between two labels. section 2.4.2
Compared with binary classification, the network ends with 10 output values. A PyTorch model uses progressively smaller hidden layers—256, 128, and 64 neurons—before producing those outputs. CrossEntropyLoss handles the logits without adding a separate softmax layer. section 2.4.2
Classifying clothing images
Flatten each 28×28 image, train the network with Adam and CrossEntropyLoss, and select the class with torch.argmax. Testing five images produces predicted labels, while evaluation across the test loader reports an accuracy of about 0.882.
