16 weekly classes from Sat Oct 31, 2026 to Sat Feb 13, 2027, then Round 1 on Fri Feb 19, 2027; Round 2 in March or April 2027 (date TBA). Starts from Python basics and covers Modules 0–5. 每周六上午上课，每次2小时。10月31日重新开课，从 Python 基础讲起，共16次课，直通2027年2月19日第一轮；第二轮在2027年3月或4月（日期待定）。
One class a week covers what used to be 2–8 short sessions, so every class mixes teaching with hands-on practice. 每周一次长课：讲解、现场演示、课堂练习交替进行。
Go over last week's homework. Each student explains one fix out loud.
2–3 ideas. For each: explain → live demo → "You try" in class. 5-minute break at the hour.
Open this week's homework together and confirm what "done" looks like.
| # | Date | Module | Topic | Lectures & homework |
|---|---|---|---|---|
| 1 | Oct 31Sat · 2 h | Module 0 | Welcome · Python I–III Road to Round 1 & Colab setup; variables, types, f-strings; if/elif/else; for/while loops. Covers old sessions 1–4 |
| # | Date | Module | Topic | Lectures & homework |
|---|---|---|---|---|
| 2 | Nov 7Sat · 2 h | Module 0 | Python IV–V · functions & data structures def/return/scope; lists, dicts, comprehensions. Checkpoint A: timed Python drill. Covers old sessions 5–7 | |
| 3 | Nov 14Sat · 2 h | Module 0 | NumPy Arrays, shape, dtype; vectorized math & broadcasting; axes, reductions, boolean masks. Covers old sessions 8–10 | |
| 4 | Nov 21Sat · 2 h | Module 0 | pandas & plotting · Module 0 mini-mock DataFrames, loc/iloc, groupby; matplotlib & seaborn; load→clean→explore→plot. 30-min mini-mock. Covers old sessions 11–16 | |
| 5 | Nov 28Sat · 2 h | Module 1 | Linear algebra I Vectors & dot products; matrices as transformations; inverse, determinant, rank. Covers old sessions 17–19 |
| # | Date | Module | Topic | Lectures & homework |
|---|---|---|---|---|
| 6 | Dec 5Sat · 2 h | Module 1 | Linear algebra II · probability & statistics Eigenvalues, SVD (idea); distributions, Bayes' rule, MLE & correlation. Covers old sessions 20–25 | |
| 7 | Dec 12Sat · 2 h | Module 1 | Calculus & gradient descent · Module 1 mock Derivatives, partials, gradient, chain rule; descend a loss by hand and in code. Timed mock. Covers old sessions 26–28 | |
| 8 | Dec 19Sat · 2 h | Module 2 | Supervised learning · linear regression Train/test split; MSE; normal equation; gradient descent; linear regression from scratch in NumPy. Covers old sessions 29–33 | Lecture coming HW 29–33 coming due Dec 26 |
| 9 | Dec 26Sat · 2 h | Module 2 | Classification · logistic regression Sigmoid & odds; cross-entropy and its gradient; logistic regression from scratch; the scikit-learn API. Covers old sessions 34–37 | Lecture coming HW 34–37 coming due Jan 2 |
| # | Date | Module | Topic | Lectures & homework |
|---|---|---|---|---|
| 10 | Jan 2Sat · 2 h | Module 2 | Model selection & classic models · Module 2 mock Bias–variance, ridge/lasso, cross-validation; k-NN, trees, ensembles, SVM; metrics. Timed mock. Covers old sessions 38–44 | Lecture coming HW 38–44 coming due Jan 9 |
| 11 | Jan 9Sat · 2 h | Module 3 | Unsupervised learning k-means (and from scratch); PCA idea and via eigen-decomposition; choosing k. Module 3 checkpoint. Covers old sessions 45–50 | Lecture coming HW 45–50 coming due Jan 16 |
| 12 | Jan 16Sat · 2 h | Module 4 | Neural networks I · backprop Perceptron & activations; MLP forward pass by hand; backpropagation by hand; softmax; SGD/momentum/Adam. Covers old sessions 51–58 | Lecture coming HW 51–58 coming due Jan 23 |
| 13 | Jan 23Sat · 2 h | Module 4 | Neural networks II · PyTorch · Module 4 mock Tensors & autograd; nn.Module & the training loop; dropout, batch norm, reading loss curves. Timed mock. Covers old sessions 59–64 | Lecture coming HW 59–64 coming due Jan 30 |
| 14 | Jan 30Sat · 2 h | Module 5 | CNNs · Module 5 mock Images as tensors; convolution & output shapes by hand; pooling; CNN in PyTorch; transfer learning. ⚠ Registration closes tomorrow (Jan 31)! Covers old sessions 65–72 | Lecture coming HW 65–72 coming due Feb 6 |
| # | Date | Module | Topic | Lectures & homework |
|---|---|---|---|---|
| 15 | Feb 6Sat · 2 h | Review | Full mock #1 debrief · Round 1 strategy Take full 3-hour mock #1 at home before class. In class: score it, weak-spot list, patch math & coding, past problems. Covers old sessions 73–79 | HW 73–79 coming due Feb 13 |
| 16 | Feb 13Sat · 2 h | Review | Full mock #2 debrief · final review Take full mock #2 at home before class. In class: debrief, speed drills, one-page formula sheet, exam-day logistics. Covers old sessions 80–84 | Lecture coming HW 80–84 coming before Round 1 |
| 🎯 | Feb 19Fri | Round 1 | ROUND 1 — competition day 12:00–3:00 pm ET (Taipei: Sat Feb 20, 1–4 am). Part 1 non-coding 60 min; Part 2 coding 75 min. Proctored, Google Colab. | Calendar |
| # | Date | Module | Topic | Lectures & homework |
|---|---|---|---|---|
| 🎯 | Mar / Aprdate TBA | Round 2 | ROUND 2 — for Round 1 qualifiers March or April 2027 — exact date to be announced on usaaio.org. Same format as Round 1, except some problems may require GPUs. 第二轮：2027年3月或4月（具体日期待官方公布）。形式与第一轮相同，但部分题目可能需要使用 GPU。 | usaaio.org |