Neural Network Architectures

Covers canonical architectures (CNNs, RNNs, Transformers) and design patterns such as residual connections, attention mechanisms, and normalization layers. Focuses on architecture choices that match data modality and compute constraints.