This course is targeted at graduate students who are interested in learning about intelligent agents capable of automating the processes of perceiving and representing the state of the world, making decisions and learning, and coordinating and competing with each other.
Techniques from Probability, Statistics, Economics, Algorithms, Operations Research and Optimal Control are increasingly important tools for improving the intelligence and autonomy of machines, whether those machines are robots traversing our environments, schedulers moving inventory, spacecraft deciding which experiments to perform, or agents finding matches of kidney donors to patients. The course will cover the ideas underlying these techniques.
The course evaluation will include multiple programming assignments, so the students should be able to program in a language of their choice. Recommended languages are C, C++, Java, Python, or Ruby.
The course textbook is Artifical Intelligence: A Modern Approach, 3rd edition, by Russell and Norvig. This is not a required purchase. We will supplement readings in this book with timely research papers posted to the course website.
The course is taught by professors Tuomas Sandholm (CSD) and Manuela Veloso (CSD). The teaching assistants are John Dickerson (CSD) and Prateek Tandon (RI). The course is open to graduate students in the School of Computer Science; interested and qualified undergraduates and other students should contact the professors for permission to join.
We meet most Mondays and Wednesdays from 10:30am to 11:50am in GHC 4303. The first lecture will be held on Monday, January 14. Check the lecture schedule below for details!
The TAs will also hold weekly informal recitations at 1:00pm on (some) Fridays in NSH 3002. These will be announced in class and over the mailing list. Weekly office hours will be held at the following times and locations:
|John Dickerson||Tuesday||2:00–4:00pm||GHC 9219|
|Tuomas Sandholm||By appointment||By appointment||GHC 9205|
|Prateek Tandon||Thursday||2:00–4:00pm||NSH 4215|
|Manuela Veloso||By appointment||By appointment||GHC 7002|
All lecture and homework dates and topics are subject to change. This is a rough outline of the topics we will be covering this semester.
|5/6/2013||Dropbox||Final Papers Due||Everyone!||—||by 11:59pm|
|5/3/2013||7th floor atrium||Poster presentations||Everyone!||—||10:00am–1:00pm|
|5/1/2013||GHC 4303||Q&A and open problems in AI||Sandholm, Veloso||—||—|
|4/29/2013||GHC 4303||Game solving (algorithms for tree games of incomplete information)||Sandholm||(pptx)|
|4/24/2013||GHC 4303||Game solving (algorithms for tree games of incomplete information)||Sandholm||(pptx)|
|4/22/2013||GHC 4303||Game solving (algorithms for tree games of incomplete information)||Ganzfried||—|
|4/17/2013||GHC 4303||Game solving (algorithms for sequential complete-information games)||Sandholm||(ppt)|
|4/15/2013||GHC 4303||State estimation (Kalman filters, particle filters, learning from demonstration), discussion of HW3 solutions||Dickerson, Tandon||HW4 out!|
|4/10/2013||GHC 4303||Opponent modeling and learning||Veloso||—||Project milestones due|
|4/8/2013||GHC 4303||Learning to win – thresholded rewards||Veloso||HW3 due!|
|4/3/2013||GHC 4303||Robot motion planning||Veloso||(IROS-2002), (NRL-2005)|
|4/1/2013||GHC 4303||Bayes Nets: Representation and Inference||Veloso||—|
|3/27/2013||GHC 4303||Game solving (algorithms for normal-form games)||Sandholm||(ppt)|
|3/25/2013||GHC 4303||Game solving (game classes, representations, solution concepts)||Sandholm||(ppt), HW3 out!|
|3/20/2013||GHC 4303||Bayesian networks, POMDPs||Veloso||—||AIMA Ch. 17|
|3/18/2013||GHC 4303||Reinforcement learning||Veloso||Project proposals due, (reading)|
|3/6/2013||GHC 4303||Midterm exam||—||(solutions)|
|3/4/2013||GHC 4303||Discussion of HW1 and HW2 solutions; review||Dickerson, Tandon||(ppt)|
|2/27/2013||GHC 4303||Planning under uncertainty, MDPs||Veloso||AIMA Ch. 17, HW2 due!|
|2/25/2013||GHC 4303||Heuristics and uncertainty in planning||Veloso||AIMA Ch. 10|
|2/20/2013||GHC 4303||Planning||Veloso||AIMA Ch. 10|
|2/18/2013||GHC 4303||Planning||Veloso||AIMA Ch. 10|
|2/13/2013||GHC 4303||Advanced informed search, MIP||Sandholm||(ppt) HW1 due, HW2 out!|
|2/11/2013||GHC 4303||Advanced informed search, MIP||Sandholm||(ppt)|
|2/6/2013||GHC 4303||Advanced informed search, MIP||Sandholm||(ppt, ppt)|
|2/4/2013||GHC 4303||Informed search||Sandholm||(ppt)|
|1/30/2013||GHC 4303||Uninformed search, SAT, CSP||Sandholm||(ppt)|
|1/28/2013||GHC 4303||Uninformed search, SAT, CSP||Sandholm||(ppt) HW1 out!|
|1/23/2013||GHC 4303||Uninformed search, SAT, CSP||Sandholm||(ppt)|
|1/21/2013||GHC 4303||Uninformed search, SAT, CSP||Sandholm||AIMA Ch. 3, (ppt)|
|1/16/2013||GHC 4303||KR for agent architectures, inference||Veloso||AIMA Ch. 7|
|1/14/2013||GHC 4303||Introduction and agent architectures, AI and robotics||Veloso||—|
Homeworks are due at the begining of class, unless otherwise specified. You will be allowed 8 total late days without penalty for the entire semester. You may use a maximum of 3 late days per individual homework assignment. Each late day corresponds to 24 hours or part thereof. Once those days are used, you will not receive any credit for late homework. You must turn in all of the homeworks, even if for zero credit, in order to pass the course.
|#1||1/28/2013||2/13/2013||KR, SAT, CSP, uninformed search||Recitation: (ppt, pdf)||(pdf)|
|#2||2/13/2013||2/27/2013||Informed search, planning||(pdf), (code)|
|#3||3/25/2013||4/8/2013||MDPs, Q-learning, POMDPs||(32-bit, 64-bit)||(pdf)|
|#4||4/15/2013||4/29/2013||Machine learning, game theory||—||—|
In lieu of a final exam, students will complete a course project. We encourage students to combine techniques from AI with their own research for these projects. Projects will be accompanied by a 6-8 page paper due at the end of the semester and a poster presentation session on May 3 from 10am–1pm in the GHC 7th floor atrium.
Project proposals are due on March 18, and should consist of a short (2–3 pages) but well-researched summary of your project idea accompanied by a plan of execution. Students are allowed (and encouraged) to work in groups; however, the expectations we will have for your project rise proportional to the group size! We'll post some project ideas in the future.
As an example of a reasonable project proposal, please take a look at the following example from a previous Graduate AI (pdf). The proposal begins with motivation and a brief explanation of the problem, and then sets a series of goals. If the project is harder than expected, only the "75%" goals will be completed; if it's difficult level is as expected, the "100%" goals will be completed; finally, if it's easier than expected, the student plans to complete the ambitious "125%" goals as well. The more you plan now, the easier it will be to complete your project well and on time!
Sample proposal: (pdf)
One month after proposing your project, the TAs will want to check up on your progress. On April 10, you will turn in a short (3–4) page description of the work you've accomplished on your project so far. You can also discuss any changes you've made to your project plans in this document.
Sample milestone: (pdf)
On May 3 from 10am–1pm in the GHC 7th floor atrium, student groups will give short poster presentations of their projects. After these presentations (by 11:59pm on Monday, May 6), a conference-sized (10–15 single-column pages, ACM format) paper covering the project material will be due.
Grades are based on Class Participation (10%), Homeworks (40%), Final Project (30%) and the Midterm (20%).
Interested students should first register, then fill out an audit form and have one of the instructors sign it. Auditors are required to complete a class project, but no homeworks or exams: that way they can choose to focus their efforts on whichever area of AI most interests them.
Feel free to use the slides and materials available online here! If you use our slides, an appropriate attribution is requested. Please email the instructors with any corrections or improvements.