35. Deep learning for tabular data

Model numerical and categorical columns with embeddings, multilayer networks, attention-based architectures, and strong regularization. Compare neural models against tree-based baselines on realistic structured datasets.