Videos
Click on a date to show the videos for that class.
Wednesday, 8/28
Friday, 8/30
Wednesday, 9/4
Monday, 9/9
Wednesday, 9/11
Monday, 9/16
Wednesday, 9/18
(Thanks to Professor Mike Genesereth from Stanford)
Wednesday, 9/24
(Only the first 16 minutes are relevant)
Monday, 9/30
Wednesday, 10/1
Monday, 10/7
Wednesday, 10/9
These videos are supplemental. It's good to have a feel for how this is done, but I won't expect you to derive the backpropagation algorithm on a test.
Friday, 10/11
Monday, 10/21
Friday, 11/1
This video is from Pieter Abbeel's AI course at Berkely. The schedule is:
| 0-9:40 | Introduction to Markov Models. WATCH THIS. |
| 9:40 -25:00 | Stationary Distributions of Markov Models. You don't need to watch this, but it is interesting. He explains Google's PageRank algorithm in the context of Markov Models. |
| 25:00-27:30 | Gibbs Sampling. Skip this. |
| 27:30-47:00 | Introduction to Hidden Markov Models. WATCH THIS. |
| 47:00-END | Inference in Hidden Markov Models. We'll do this on Monday. |
Monday, 11/4
This video is from Pieter Abbeel's AI course at Berkely. The schedule is:
| 0-9:40 | Introduction to Markov Models. ALREADY WATCHED. |
| 9:40 -25:00 | Stationary Distributions of Markov Models. You don't need to watch this, but it is interesting. He explains Google's PageRank algorithm in the context of Markov Models. |
| 25:00-27:30 | Gibbs Sampling. Skip this. |
| 27:30-47:00 | Introduction to Hidden Markov Models. ALREADY WATCHED. |
| 47:00-END | Inference in Hidden Markov Models. WATCH THIS. |
This video is the sequel to the previous video:
| 0-9:00 | HMM Review. SKIP THIS. |
| 9:00 -26:50 | Introduction to particle filters. WATCH THIS. |
| 26:50-35:00 | Particle Filtering Quiz. SKIP THIS. |
| 35:00-40:00 | Example of using a particle filter for robot localization. WATCH THIS IF YOU WANT. |
| 40:00-50:00 | Simultaneous Localization and Mapping (SLAM). SKIP THIS. |
| 50:00-56:00 | Dynamic Bayes Nets. WATCH THIS. |
| 56:00-END | More application examples. Watch if you want. |
Wednesday, 11/6
Another video from Pieter Abbeel's CS188 course at Berkely. If you have time, you should watch the whole thing. You should definitely watch:
| 0-13:30 | Introduction to Markov Decision Processes. |
| 22:40 -29:00 | Assigning values to reward sequences. |
| 41:21-101:30 | Value iteration. |
Monday, 11/11
Another video from Pieter Abbeel's CS188 course at Berkely.
Monday, 11/17
This is the final in our series of 188 video. Here is a rough schedule:
| 0-13:20 | Review of MDP's and Q-learning. |
| 13:20-20:20 | Exploration vs. Exploitation. WATCH THIS. It relates to PA4. |
| 20:20-30:50 | Exploration function and regret. This introduces a more efficient way to handle exploration. It is interesting, but not required. |
| 31:00-50:50 | Approximate Q-learning. WATCH THIS. This describes the idea behind the final question for the PA. |
| 50:50 - 59:00 | Justification/Explanation of approximate Q-learning in terms of linear regression. This is interesting, and ties nicely back to material from earlier in the course, but it isn't required material. |
| 59:00 - END | Policy search for reinforcement learning. A different approach to RL. This relates to Friday's discussion of heuristics for function optimization. Again, this is is interesting but not required. |