DCS 520 -- Graduate Introduction to AI

Spring 2001

Main Page Contents
Schedule
Lectures
Materials
Links

Course People
Matthew Stone

Schedule

Class MW6 (4:30-5:50) Hill 120
Office Hours W 2:00-4:00 CoRE 328

Announcements

  • Apr 18.
    The final exam is officially scheduled for Monday, May 7 at 4pm, in Hill 120 (the normal class room).

  • Apr 17.
    A sample executable for the learner program is available on paul as /grad/u1/mdstone/520/learner and on romulus/remus as ~mdstone/520/learner (Apr 19). So you can play to see what one satisfactory implementation does.

  • Apr 16.
    I have made a slight modification to router.tcl to make your assignments easier to debug. If you want to print out information as your program runs with the tcl simulation visualizer, you can use fprintf(stderr...). This output will be redirected to the terminal of the tcl simulation visualizer, and by default will be printed there. (FYI, the tcl simulation visualizer expects to read and interpret the output that the simulator writes to stdout, so you won't see any additional information you write to stdout, and what's more it will probably cause an error.)

  • Apr 11.
    Materials for the last part of class, on planning under uncertainty and on evaluation, are on reserve in the math library.

    Here's that homework I promised, due April 26.

  • Mar 26.
    I have updated the syllabus to reflect what we have actually done, and what we can do in the time that remains. The decision model notes are now a complete draft. I have provided a text resource for Kalman filters. I have also updated the research papers list with a few influential current applications of probabilistic inference for dynamic environments, for HMMs and related material, to complement the natural language applications described by Charniak.

    I have now posted the final assignment for the class. (Final in one sense, because it's due at the end, on April 30; not final in another sense, because there will be more homework.) This is a written assignment, and you have plenty of time to think about it, research it and write it up. Take advantage of the time!

  • Mar 5. Snow!
    No class today. Hand in homework Wednesday Mar 7. Midterm Mar 7 as always.

  • Feb 19.
    A second round of written and programming exercies now available, due Mar 5.
    Code to run experiments (see assignment).

  • Feb 12.
    A first round of written exercies now available, due Feb 21.

  • Jan 31.
    Partial draft of Decision model notes now available.
    No class Feb 5.

Lecture Schedule, AI Events, Notes

  • Jan 17
    What is AI?

  • Jan 22
    Decision analysis as a computational model (1)
    Actions, observations, outcomes; interpreting models; solving for and carrying out policies in agents.

  • Jan 24
    Decision analysis as a computational model (2)
    Training data; model estimation and model induction; generalization, model selection, data sparsity.

  • Jan 29
    Decision analysis as a computational model (3)
    Design and evaluation; sensitivity analysis, computational complexity, statistical hypotheses about running agents.

  • Jan 31
    Bayesian analysis and models for classification (1)
    Naive Bayes inference. Text classification.

  • Feb 5
    No class.

  • Feb 7
    Bayesian analysis and models for classification (2)
    Continuous variables, normal distributions, linear classifiers.

  • Feb 12
    Bayesian analysis and models for classification (3)
    Learning from training data. Maximum likelihood.

  • Feb 13
    Sebastian Thrun visit to RuCCS.

  • Feb 14
    Bayesian analysis and models for classification (4)
    General density estimation: nearest neighbor classification; Parzen windows.

  • Feb 19
    Bayesian analysis and models for classification (5)

  • Feb 21
    Bayesian analysis and models for classification (6)
    Clustering; k-means; expectation maximization.

  • Feb 26 Models of time (1)
    Markov models and hidden Markov models.

  • Feb 28 Models of time (2)
    HMM decoding. Part-of-speech tagging.

  • Mar 5
    Snow.

  • Mar 7
    Midterm.

  • Mar 19
    Models of time (3)
    HMM training. Forward/backward. Speech and gesture recognition.

  • Mar 21
    Models of time (4)
    The Kalman filter. Tracking and learning with Gaussian priors.

  • Mar 26
    Models of hierarchical structure. Trees, CFGs and PCFGs.

  • Mar 28
    Using PCFG models: inside-outside algorithm and PCFG parsing.

  • Apr 2
    General probabilistic inference: belief nets and influence diagrams.

  • Apr 4
    Multiple decisions and the value of information. Active perception.

  • Apr 9
    Markov decision processes: Value iteration.

  • Apr 11
    Markov decision processes: Policy iteration.

  • Apr 16
    Reinforcement learning.

  • Apr 18
    Evaluation (1).
    Pitfalls and methodology.

  • Apr 23
    Evaluation (2).
    Performance metrics.

  • Apr 25
    Evaluation (3).
    Training and test data. Reliability. Cross-validation.

  • Apr 30 Slack.

  • May 7 Final, 4-5:30.

Materials

Links