IITM BS-DS · DL GEN AI · SHARED LESSONS
Seven lessons, newest first. Each page carries a drill you can attempt without notes.
| Lesson | Covers |
|---|---|
| Week 4 — U-Net | Segmentation, contracting and expanding paths, skip connections, training and evaluation. |
| Week 4 — transfer learning | Pretrained features, input normalization, frozen bodies, fine-tuning, and the food-classifier walkthrough. |
| Week 4 — famous CNNs | LeNet to ResNet: architecture changes, parameter budgets, and residual paths. |
| Week 3 — CNN build | Convolution and pooling from scratch, PyTorch layers, dataloaders, and a training loop. |
| Week 3 — CNN theory | Receptive fields, weight sharing, shapes, parameters, layers, and architecture. |
| Week 2 — MLP in PyTorch | Churn and FashionMNIST, initialization, training, validation, and metrics. |
| Week 2 — MLP training theory | Universal approximation, initialization, gradient descent, backpropagation, and XOR. |