Statistics 154/254: Statistical Machine Learning
UC Berkeley, Spring 2025
Course Content and Calendar
This course is an introduction to the statistical concepts that underpin our understanding of modern machine learning.
The core concepts are:
- A taxonomy of ML tasks
- Inference versus prediction
- Regression models
- Classification models
- Unsupervised learning
- Loss minimization
- Task–appropriate loss functions
- Generalization from empirical risk to population risk
- Estimating risk in practice
- Complexity
- How to form classes of expressive models
- Implicit and explicit regulariziation
- The costs and benefits of complexity (e.g., the bias / variance tradeoff)
- Computation and estimation
- Black-box optimization
- Stochastic optimization
- Automatic differentiation
The following schedule is aspirational and subject to change as we go.
Week | Date | Day | Note | Unit | Topic | Reading | Assignment |
---|---|---|---|---|---|---|---|
1 | Jan 22 | Wednesday | Unit 0: Introduction and review | Course policies and introduction | |||
1 | Jan 24 | Friday | What statistical prediction is and isn’t | ||||
2 | Jan 27 | Monday | Unit 1: Regression | Population loss minimization | |||
2 | Jan 29 | Wednesday | Linear regression as empirical loss minimization | ||||
2 | Jan 31 | Friday | Making new features out of old | HW0 due | |||
3 | Feb 3 | Monday | Bias / variance tradeoff with feature selection | ||||
3 | Feb 5 | Wednesday | L2 penalization | ||||
3 | Feb 7 | Friday | Guest or recorded lecture | L1 penalization | |||
4 | Feb 10 | Monday | Unit 2: Risk and complexity | Uniform laws and generalization error | |||
4 | Feb 12 | Wednesday | A uniform law for smooth functions | ||||
4 | Feb 14 | Friday | VC dimension, zero–one loss, and generalization | HW1 due | |||
5 | Feb 17 | Monday | Administrative holiday | ||||
5 | Feb 19 | Wednesday | (missed due to sickness) | ||||
5 | Feb 21 | Friday | Review and quiz | Quiz1 | |||
6 | Feb 24 | Monday | Cross validation and held-out sets | ||||
6 | Feb 26 | Wednesday | Cross validation for model selection | ||||
6 | Feb 28 | Friday | Homework Q&A | HW2 due | |||
7 | Mar 3 | Monday | Unit 3: Classification | Classification loss and proxy loss functions | |||
7 | Mar 5 | Wednesday | Discriminative and generative losses | ||||
7 | Mar 7 | Friday | Review and quiz | Quiz2 | |||
8 | Mar 10 | Monday | Guest or recorded lecture | ROC curves for classification | |||
8 | Mar 12 | Wednesday | Guest or recorded lecture | The perceptron algorithm | |||
8 | Mar 14 | Friday | Guest or recorded lecture | Support vector (max-margin) classifiers | (no quiz or HW) | ||
9 | Mar 17 | Monday | Floating unit: Optimization | Gradient descent | |||
9 | Mar 19 | Wednesday | Stochastic gradient descent | ||||
9 | Mar 21 | Friday | Homework Q&A | HW3 due | |||
10 | Mar 24 | Monday | Spring Break | ||||
10 | Mar 25 | Tuesday | Spring Break | ||||
10 | Mar 26 | Wednesday | Spring Break | ||||
10 | Mar 27 | Thursday | Spring Break | ||||
10 | Mar 28 | Friday | Spring Break | ||||
11 | Mar 31 | Monday | Canceled class | Unit 4: Trees and weak learners | No lecture | ||
11 | Apr 2 | Wednesday | Regression and classification trees | ||||
11 | Apr 4 | Friday | Review and quiz | Quiz 3 | |||
12 | Apr 7 | Monday | Bagging | ||||
12 | Apr 9 | Wednesday | Boosting | ||||
12 | Apr 11 | Friday | Review and quiz | Quiz X (Review) | |||
13 | Apr 14 | Monday | Unit 5: Kernels and interpolators | Inner product spaces and the polynomial kernel | |||
13 | Apr 16 | Wednesday | The kernel trick in SVMs and ridge regression | ||||
13 | Apr 18 | Friday | Positive definite kernels | HW 4 due | |||
14 | Apr 21 | Monday | Reproducing kernel Hilbert spaces | ||||
14 | Apr 23 | Wednesday | The representer theorem and interpreting kernels | ||||
14 | Apr 25 | Friday | Review and quiz | Quiz 4 | |||
15 | Apr 28 | Monday | Class review | Class review | |||
15 | Apr 30 | Wednesday | Class review | ||||
15 | May 2 | Friday | Class review | HW5 due | |||