Calendar

Fall 2026

The topics and dates of future lectures may change.

Fall quarter runs from Wednesday, September 23 to Friday, December 11, 2026. Class meets on Mondays, Tuesdays, Wednesdays and Fridays.

Homework is due on Canvas at 11:59 pm on its due date. Each self-grade is due one week after that homework’s solutions are posted; the self-grade dates below assume the solutions go up at the deadline. Homework 1 and Homework 2 are each due on a Thursday and stay open with no penalty until 11:59 pm on Friday, so their solutions go up after Friday’s deadline and each self-grade is due one week later. Homework 3 is due on Wednesday, November 4, at 11:59 pm, and stays open with no penalty until 11:59 pm on Thursday, November 5. Its solutions go up on Friday, November 6, a week before Midterm 2. There is no class on Thursdays; the table’s Thursday column shows only deadlines and holidays. Homework 4 is due on Wednesday, November 18, at 11:59 pm, and stays open with no penalty until 11:59 pm on Thursday, November 19. Its self-grade is due Tuesday, December 1, since the week after its solutions go up is the Thanksgiving break. Homework 5 is due on Friday, December 4, at 11:59 pm, and stays open with no penalty until 11:59 pm on Saturday, December 5. Its solutions go up on Sunday, December 6, and its self-grade is due Friday, December 11, not a week later, because WWU’s fall grades are due on Tuesday, December 15. Homework 1 to 5 are the course’s five homework assignments. Homework 3 was first announced as Homework 2.5.

The final exam is optional. It is on Friday, December 11, 8–10am. WWU’s finals schedule sets each exam’s slot from the class’s start time and meeting days; this class starts at noon and meets MTWF, which WWU lists among classes that include a Tuesday or Thursday meeting.

Midterms 1, 2 and 3 are in class on Friday, October 30, Friday, November 13 and Monday, November 23. Midterm 2 moved from Monday, November 9 on October 27, after gradient descent moved to Monday, November 2. Midterm 1 covers linear regression and generalization. Midterm 2 covers gradient descent, logistic regression and performance metrics. Neither covers regularization or model selection. Midterm 3 covers the lectures that neither of them covers, from regularization through the introduction to backpropagation, including the support vector machine lectures of November 6 to November 16. The lectures after Midterm 3 (backpropagation, minibatched training, ML experimentation, clustering and ethics) are not on a midterm, only on the optional final. Midterm 1 has an in-class review session on Wednesday, October 28. Midterm 2 has one on Tuesday, November 10, and Midterm 3 has one on the Friday before the exam. The final’s review session will be held during finals week, at a time TBD.

	Monday	Tuesday	Wednesday	Thursday	Friday
Week 0 Sep 20 – Sep 26			Syllabus		Course overview (begun)
Week 1 Sep 27 – Oct 3	Course overview	Course overview	Course overview Linear regression (begun)		Linear regression
Week 2 Oct 4 – Oct 10	Math refreshers (begun)	Math refreshers	Linear regression		Linear regression lab
Week 3 Oct 11 – Oct 17	Generalization: risk, held-out estimates, cross-validation	Generalization: risk, held-out estimates, cross-validation	Generalization: bias, variance and residual variance	Homework 1 due 11:59 pm (no penalty until Fri, Oct 16, 11:59 pm)	Generalization: bias, variance and residual variance
Week 4 Oct 18 – Oct 24	Classification overview Loss functions for classification	Maximum likelihood estimation	The Bernoulli MLE	Homework 2 due 11:59 pm (no penalty until Fri, Oct 23, 11:59 pm)	Logistic regression: the logistic and logit functions Homework 1 self-grade due
Week 5 Oct 25 – Oct 31	Logistic regression: cross-entropy, its gradient, why there is no formula for the fit	Logistic regression: Newton’s method	Midterm 1 review		Midterm 1 Homework 2 self-grade due
Week 6 Nov 1 – Nov 7	Gradient descent Logistic regression: fitting on real data, when the coefficients diverge Logistic regression: what the coefficients mean, where the boundary is	Performance metrics	Regularization: ridge regression, the lasso Model selection: tuning a hyperparameter by cross-validation Homework 3 due 11:59 pm (no penalty until Thu, Nov 5, 11:59 pm)		Hard-margin SVM: the primal problem
Week 7 Nov 8 – Nov 14	Hard-margin SVM: the dual problem	Midterm 2 review	No class: Veterans Day	Homework 3 self-grade due	Midterm 2
Week 8 Nov 15 – Nov 21	Soft-margin SVMs: slack, \(C\) and the hinge loss Non-linear SVMs: the kernel trick	Introduction to neural networks: nonlinear prediction, learned features Layers, and the representational power of neural networks	Deep neural networks Introduction to backpropagation Homework 4 due 11:59 pm (no penalty until Thu, Nov 19, 11:59 pm)		Midterm 3 review
Week 9 Nov 22 – Nov 28	Midterm 3	Backpropagation	No class: Thanksgiving	Thanksgiving	No class: Thanksgiving
Week 10 Nov 29 – Dec 5	Minibatched backpropagation Minibatch stochastic gradient descent	ML experimentation: the training loop, splits, preprocessing, model families, baselines, ensembling Homework 4 self-grade due	Clustering overview K-means clustering		Ethics and fairness Course evaluations Homework 5 due 11:59 pm (no penalty until Sat, Dec 5, 11:59 pm)
Finals week Dec 7 – Dec 11	Optional final exam Friday, Dec 11, 8–10am, per the WWU finals schedule Review session: during finals week, time TBD Homework 5 self-grade due: Friday, Dec 11

1 Optional reading

Lectures this term skip these parts of the notes. They stay in the notes as optional reading, and the study guide’s outcomes about them are optional too.

• Logistic regression: more than two classes. The softmax is the exception: neural networks use it as their output layer, so work the softmax exercise before the neural network lectures in week 8.
• Soft-margin SVMs, from the soft-margin dual to the \(\svfrac\)-SVM: deriving the dual, what its \(\dualvar_i\) say about each training point, and the \(\svfrac\)-SVM. The kernel-trick lecture states the soft-margin dual where it needs it.
• Multiclass SVMs (one-vs-one and one-vs-rest).
• ML theory: the No Free Lunch theorem and the Ugly Duckling theorem.
• Unsupervised learning, apart from the clustering overview and k-means clustering.
• Ensemble methods and Non-parametric models, which this term’s lectures skip.
• ML theory: Bayesian decision theory and the curse of dimensionality. Reinforcement learning, simulated annealing and genetic algorithms are not lectured either.
• Regularization: regularizing without a penalty, on data augmentation.
• Two chapters that no lecture covers: Semi-supervised learning and Attention and transformers.