Course Schedule
Fall 2026 · Tuesday & Thursday, 12:30–1:45 p.m.
Thirty-one class meetings run from Tuesday, August 25 through Thursday, December 10, 2026. Chapter numbers refer to Prince, Understanding Deep Learning (MIT Press, 2024).
Read the assigned chapter before the first meeting of the week. Lecture time is spent on the parts that are hard to get from reading alone: derivations, failure modes, and code.
Weekly assignments are posted in Blackboard with their own due dates; the Due column below carries only the exam and project milestones.
Weekly Plan
| Week | Dates | Topic | Reading | Due |
|---|---|---|---|---|
| 1 | Aug 25, 27 | Introduction; supervised learning | Ch. 1–2 | — |
| 2 | Sep 1, 3 | Mathematical & Python background | App. B | — |
| 3 | Sep 8, 10 | Shallow neural networks | Ch. 3 | — |
| 4 | Sep 15, 17 | Deep neural networks | Ch. 4 | — |
| 5 | Sep 22, 24 | Loss functions | Ch. 5 | — |
| 6 | Sep 29, Oct 1 | Fitting models | Ch. 6 | — |
| 7 | Oct 6, 8 | Gradients & initialization | Ch. 7 | Proposal |
| 8 | Oct 13, 15 | Measuring performance | Ch. 8 | — |
| 9 | Oct 20, 22 | Midterm review; midterm Oct 22 | Ch. 1–7 | Midterm |
| 10 | Oct 27, 29 | Regularization; convolutional networks | Ch. 9–10 | — |
| 11 | Nov 3, 5 | Residual networks; transformers | Ch. 11–12 | Paper choice |
| 12 | Nov 10, 12 | Large language models | Ch. 12 + notes | Progress |
| 13 | Nov 17, 19 | Graph neural networks | Ch. 13 | — |
| 14 | Nov 24 | VAEs & GANs (no class Nov 26) | Ch. 15, 17 | — |
| 15 | Dec 1, 3 | Diffusion models; presentations begin | Ch. 18 | — |
| 16 | Dec 8, 10 | Student presentations | — | Report Dec 14 |
Meeting-by-Meeting Detail
Part I — Foundations
| Date | Session | Topic |
|---|---|---|
| Tue, Aug 25 | 1 | Course overview; what deep learning is and is not |
| Thu, Aug 27 | 2 | Supervised learning: models, parameters, training |
| Tue, Sep 1 | 3 | Linear algebra and calculus you actually need |
| Thu, Sep 3 | 4 | Probability refresher; PyTorch tensors and autograd |
| Tue, Sep 8 | 5 | Shallow networks: one hidden layer, ReLU, linear regions |
| Thu, Sep 10 | 6 | Universal approximation; shallow networks in PyTorch |
| Tue, Sep 15 | 7 | Composing networks; matrix notation for deep models |
| Thu, Sep 17 | 8 | Depth vs. width; counting linear regions |
Part II — Training Neural Networks
| Date | Session | Topic |
|---|---|---|
| Tue, Sep 22 | 9 | Maximum likelihood; recipe for constructing losses |
| Thu, Sep 24 | 10 | Regression, binary and multiclass classification losses |
| Tue, Sep 29 | 11 | Gradient descent; loss surfaces and local minima |
| Thu, Oct 1 | 12 | Stochastic gradient descent, momentum, Adam |
| Tue, Oct 6 | 13 | Backpropagation derived and implemented from scratch |
| Thu, Oct 8 | 14 | Vanishing/exploding gradients; He and Glorot initialization |
| Tue, Oct 13 | 15 | Training, validation, and test error; the bias–variance trade-off |
| Thu, Oct 15 | 16 | Double descent; hyperparameter search; midterm review |
| Tue, Oct 20 | 17 | Midterm review |
| Thu, Oct 22 | 18 | Midterm exam |
| Tue, Oct 27 | 19 | Regularization: explicit and implicit; early stopping, ensembles, dropout, noise, transfer learning |
Midterm Exam — Thursday, October 22
In class, multiple-choice, closed book, covering Chapters 1–7.
Part III — Architectures
| Date | Session | Topic |
|---|---|---|
| Thu, Oct 29 | 20 | Convolution, pooling, and image classification |
| Tue, Nov 3 | 21 | Residual connections, batch norm, and modern CNNs |
| Thu, Nov 5 | 22 | Self-attention and the transformer block; positional encoding; encoder, decoder, and encoder–decoder models |
| Tue, Nov 10 | 23 | Pretraining, fine-tuning, and scaling laws |
| Thu, Nov 12 | 24 | Tokenization, RoPE, KV caching, and inference efficiency |
| Tue, Nov 17 | 25 | Graph representation; permutation equivariance |
| Thu, Nov 19 | 26 | Graph convolutions; node, edge, and graph-level tasks |
Part IV — Generative Models & Presentations
| Date | Session | Topic |
|---|---|---|
| Tue, Nov 24 | 27 | Latent variable models: VAEs and GANs |
| Thu, Nov 26 | — | No class — Thanksgiving holiday |
| Tue, Dec 1 | 28 | Denoising diffusion: forward process, reverse process, sampling |
| Thu, Dec 3 | 29 | Student presentations |
| Tue, Dec 8 | 30 | Student presentations |
| Thu, Dec 10 | 31 | Student presentations; course wrap-up |
NDSU Academic Calendar Dates
Drawn from the NDSU Fall 2026 academic calendar.
| Date | Event |
|---|---|
| Mon, Aug 24 | Classes begin at 4:00 p.m. |
| Tue, Aug 25 | First full day of classes |
| Wed, Sep 2 | Last day to add; last day for no-record drop at 100% refund |
| Mon, Sep 7 | Labor Day — no classes |
| Mon, Sep 14 | Last day to request audit or pass/fail |
| Mon, Sep 21 | Enrollment census |
| Fri, Oct 16 | Grades of ‘Incomplete’ convert to ‘F’ |
| Wed, Nov 11 | Veterans Day — no classes |
| Fri, Nov 13 | Last day to drop with a ‘W’ record |
| Wed–Fri, Nov 25–27 | Thanksgiving holiday — no classes |
| Mon–Fri, Dec 7–11 | Dead Week |
| Fri, Dec 11 | Last day of Fall classes |
| Mon–Fri, Dec 14–18 | Final examinations |
Labor Day and Veterans Day both fall on non-class days this semester, so Thanksgiving is the only holiday that removes a meeting from our Tuesday/Thursday slot.