CSE 250A. Principles of Artificial Intelligence:
Probabilistic Reasoning and Decision-Making
| Administrivia | Syllabus | Piazza | GradeSource |
Subject
Probabilistic methods for reasoning and decision-making under uncertainty. Topics include: inference and learning in directed probabilistic graphical models; prediction and planning in Markov decision processes; applications to computer vision, robotics, speech recognition, natural language processing, and information retrieval.
Prerequisites
The course is aimed broadly at advanced undergraduates and beginning graduate students in mathematics, science, and engineering. Prerequisites are elementary probability, multivariable calculus, linear algebra, and basic programming ability in some high-level language such as C, Java, or Matlab. Programming assignments are completed in the language of the student's choice.
Relation to other courses
CSE 250a covers largely the same topics as CSE 150 (as I teach it), but at a faster pace and more advanced mathematical level. The homework assignments and exams in CSE 250A are also longer and more challenging. In general you should not take CSE 250a if you already have taken CSE 150 from me in a previous quarter.
Administrivia
- Instructor: Lawrence Saul
- Teaching assistants (TAs):
Kristjan Jonsson (kjonsson
eng.ucsd.edu)
Chaitanya (Chay) Ryali (rckrishneng.ucsd.edu)
Amanda Song (mas065ucsd.edu)
- Lectures: Tue/Thu 12:30-1:50 pm, PCYNH 109
- TA sections:
Chay: Fri 3-4 pm, CSB 004
Amanda: Fri 4-5 pm, CSB 004
Kristjan: Mon 5-6 pm, WLH 2113 - Instructor office hour: Thu 3-4 pm @ EBU-3B 3214
- TA office hours:
Chay: Thu 11 am-noon, EBU-3B B250a
Amanda: Fri 5-6 pm, CSB 004
- Grading: homework (25%), two quizzes (40%), final exam (35%).
Textbooks
The course does not closely follow a particular text; the lectures are meant to be self-contained. Nevertheless, the following texts (though not required) may be useful as general references:- Artificial Intelligence: A Modern Approach (Russell & Norvig, 2010)
- Machine Learning: A Probabilistic Perspective (Murphy, 2012)
- Reinforcement Learning: An Introduction (Sutton & Barto, 1998)
- Pattern Recognition and Machine Learning (Bishop, 2006)
Syllabus
| Thu Sep 24 | Administrivia and course overview. | |
| Tue Sep 29 | Modeling uncertainty, review of probability, explaining away. | HW 1 out. |
| Thu Oct 01 | Belief networks: from probabilities to graphs. | |
| Tue Oct 06 | Conditional independence, d-separation, polytrees. | HW 1 due. HW 2 out. |
| Thu Oct 08 | Algorithms for exact and approximate inference. | |
| Tue Oct 13 | Maximum likelihood estimation; Markov models of language; naive Bayes models of text. | HW 2 due. HW 3 out. |
| Thu Oct 15 | Linear and logistic regression. Numerical optimization. | |
| Tue Oct 20 | Latent variable modeling. Expectation-Maximization (EM) algorithm. Auxiliary functions. | HW 3 due. |
| Thu Oct 22 | EM algorithm: derivation, proof of convergence. | |
| Tue Oct 27 | Quiz #1 | HW 4 out. |
| Thu Oct 29 | Examples of EM; applications to language modeling. | |
| Tue Nov 03 | Hidden Markov models, automatic speech recognition, Viterbi algorithm. | HW 4 due. HW 5 out. |
| Thu Nov 05 | Forward-backward algorithm, Gaussian mixture models, Kalman filters. | |
| Tue Nov 10 | Reinforcement learning (RL), Markov decision processes. | HW 5 due. HW 6 out. |
| Thu Nov 12 | Policy evaluation, policy improvement. | |
| Tue Nov 17 | Policy iteration, value iteration. | HW 6 due. |
| Thu Nov 19 | Stochastic approximation theory, temporal difference prediction. | |
| Tue Nov 24 | Quiz #2 | HW 7 out. |
| Thu Nov 26 | Thanksgiving Holiday | |
| Tue Dec 01 | Q-learning, extensions of RL. | |
| Thu Dec 03 | Bonus topic (if time). | HW 7 due. |
| Fri Dec 11 | Final exam. |