CS 6766: Reasoning About Uncertainty - Fall 2015
- Instructor: Joe Halpern, 414 Gates, halpern@cs.cornell.edu, 5-9562
- Admin: Randy Hess, 323 Gates, rbhess@cs.cornell.edu, 5-0985;
- TA: Matvey Soloviev, msoloviev@cs.cornell.edu
- Classes:Tuesday, Thursday 1:25 - 2:40, 362 Hollister
- Office hours
- Halpern: TBD
- Soloviev: TBD
- Text: Reasoning About Uncertainty (Halpern).
(It should be available in the bookstore. The paperback version is
marginally better (and should be cheaper), but the hardcover version is
OK too. The paperback version
corrects a number of typos and minor errors in the hardcover version.)
Unfortunately, there are still typos in paperback version; a list of
those that have been discovered so far can be found at
http://mitpress2.mit.edu/catalog/item/default.asp?ttype=2&tid=10758&xid=13&xcid=8006.
I'm sure there are even more lurking in the book; if you discover any,
let me know!
- Grading:
There will be no tests or final examination.
There will be problems handed out, typically 3 every Thursday, from the
book. The grade will be based completely on your performance on the
problems. Problems are always due two weeks after they're handed out.
If you hand them in one week after they're handed out, I will grade them
and return them the following week. You can then redo any problem that
you seriously attempted and hand it in again,
to improve your grade. On a redo, you can get a maximum of 1 point
less than the original value of the problem. (That is, if the problem
was originally out of 10, the most you can get is 9.) I will take the
higher grade.
- Academic Integrity: It's OK to discuss the problems with
others, but you MUST write up solutions on your own, and
understand what you are writing.
- Course Outline:
We will be following the text very closely. Very roughly, we will be
covering one chapter per week. Topics include approaches to
representing uncertainty, updating, representing uncertainty in
multiagent systems, logics of reasoning about uncertainty, and default
reasoning. The material should be relevant to philosophy and game
theory as well as CS; I'll try to bring out the connections as we go along.
- Webpage: The course URL is
http://www.cs.cornell.edu/courses/cs6766/2013sp. Assignments will be
posted there, as well as other class information.
- News:
- Joe's office hours will be Wednesday 3:30 - 4:30 in
Gates 414. But this week only (i.e. Sept. 2) it will be Wednesday 5-6.
- Matvey's office hours will be Mondays 2-3 (starting this week,
Aug. 31) in Gates 324.
- Piazza is now up for the course (thanks, Matvey!) and can be found
here.
- You can now hand in homework on CMS.
Homework
Note: you can find your grades and solutions to the homework on
CMS.
Week 1: handed out 8/27; due 9/10 (hand it in 9/3 for a second chance)
- Read Chapters 1, 2.1-2.3
- Do 2.5, 2.14, 2.16. In 2.5, change the second two occurrence of >=
to >.
Week 2: handed out 9/3; due 9/17 (hand it in 9/10 for a second chance)
- Read all of Chapter 2 (inclding the exercises and notes at the end).
- Think about (but don't hand in) 2.25
- Do 2.28, 2.36, 2.57(a),(c),(d).
Week 3: handed out 9/10; due 9/247 (hand it in 9/17 for a second chance)
- Read all 3.1-3.4
- Do 3.2, 3.7, 3.8
- You can now post homework on CMS!
Week 4: handed out 9/17; due 10/1 (hand it in 0/24 for a second chance)
- Read Chapter 3
- Do 3.16 (5 points), 3.23, 3.41 (5 points), 3.44
- Don't forget that you an now hand in homework on CMS.
- Steve asked if the only time that Bel|B = Bel||B iff Bel is a
probability measure. That was an interesting question! It turns out
that it's almost, but not quite true. The easiest way to think about
this is to observe Plaus(A||B) = Plaus(A \inter B)/Plaus(B), while
Plaus(A|B) = Plaus(A \inter B)/(Plaus(A \inter B) + Bel(A^c \inter B)). Thus,
Plaus|B = Plaus||B iff, for all A, either Plaus(A \inter B) = 0 or
Plaus(B) = Plaus(A \inter B) + Bel(A^c \inter B). It turns out that
this latter ocndition holds iff, for the mass function m corresponding to
Bel, either (a) the only sets that got positive mass are singletons (so
Bel is a probability measure) or (b) the only sets that get positive
mass are either singletons or the whole space S. It's not that hard to
prove this; if you're interested, I can show you the details offline.
Week 5: handed out 9/24; due 10/8 (hand it in 10/1 for a second chance)
- Read Chapter 4
- Do 3.47, 4.18, 4.24(a),(b),(c)
Week 6: handed out 10/1; due 10/15 (hand it in 10/8 for a second chance)
- Read Chapter 5
- Do 5.5, 5.8, 5.10 (There's a typo in 5.8: the subscript on E should
be calligraphic P, not \mu; there's also a typo in 5.5: there should be
a twiddle over the 0 in (b').)
Week 7: handed out 10/8; due 10/22 (hand it in 10/15 for a second chance)
- Read Chapter 6.1-6.5
- Do 5.15, 5.19, 6.4. Think about (but don't hand in) 5.38.
Week 8: handed out 10/15; due 10/29 (hand it in 10/22 for a second chance)
- Read Chapter 6
- Do 6.6 6.7, 6.14.
Week 9: handed out 10/22; due 11/4 (hand it in 10/29 for a second chance)
- Read Chapter 7
- Do 6.18, 7.11, 7.21
- Note that class is canceled for next Thursday (10/29), but homework
is still due. My office hours on 10/28 are also canceled (I'm going to
Taiwan for a few days).
Week 10: handed out 10/29; due 11/12 (hand it in 11/51 for a second chance)
Week 11: handed out 11/5 due 11/19 (hand it in 11/12 for a second chance)
- Read Chapter 8.1-8.5
- Do 8.5, 8.34 (do it only for \M^{ps}, \M^{poss}, \M^{tot}, and
\M^{qual}), 8.48
Week 12: (final homework!) handed out 10/12; due 12/1 (hand it in 10/19 for
a second chance)
- Read Chapter 10
- 8.38, 10.18, 10.24