### Announcements

*Dec 14*:
Final project writeups are due on Dec 17. Only hardcopy
will be accepted. They should placed in the box that will
be in front of my Soda Hall office door.
*Nov 29*:
The poster session will be held on Friday, 12/14 from
2:00 to 4:00 in the Wozniak Lounge in Soda Hall.
*Oct 26*:
Note that we have a Google groups site:
http://groups.google.com/group/cs281a-fall07
*Oct 26*:
The deadline for turning in Homework 4 is hereby
extended to Tue, Nov 6.
*Oct 11*:
Homework 3 is due on Tue, Oct 16. One of the GSI's will be
in the usual classroom (277 Cory) from 2:00 to 2:15 to collect
the homework. You're also welcome to turn your homework in
during Monday's recitation.
*Oct 9*:
There will be no lecture on Tue, Oct 16. (There *will* be a
lecture on Thur, Oct 11).
*Oct 9*:
There will be no lecture on Tue, Oct 16. (There *will* be a
lecture on Thur, Oct 11).
*Oct 9*:
There will be no office hours on Thu, Oct 11 and Tue, Oct 16.
Please contact me by email if you have questions.
*Sep 27*:
The recitations for Oct 8, 15, 22 and 29 will be in 2 LeConte
instead of 306 Soda.
*Sep 25*:
On homework 2, for problem 4 parts (a) and (b), please give
2x2 tables. For 4(c), by "analytic" we mean give a closed-form
expression for p(x_0 =0 | x_N = 0). This means a formula with
only simple arithmetic operations (add/subtract, multiply/divide,
exponentiate), no summations over indices nor matrix operations.
All answers should be in terms of q and maybe N and/or i. To
make the hint slightly more detailed, think of binomial coefficients (n
choose k) and, to get a closed form, consider binomial expansions
of (a+b)^n where a,b, and n are chosen appropriately. Also,
there is a small typo. When defining the edges of the Markov
chain, there should be a 1 in the set of integers from
0 to N-1. It should read 'the edges are (i, i+1) for i in
{0,1,2, ..., N-1}'.
*Sep 19*: There will be no office hours this Thursday, Sept 20th.
*Sep 13*: Percy Liang's office hours will henceforth take
place in 711 Soda Hall.
*Sep 12*: Daniel Ting says "Apologies for the confusion over
Eliminate during recitation. All the calculations/explainations for
the Eliminate algorithm given in class are correct. The time/space
complexity for the 3-clique obtained during the Eliminate algorithm
in the example given in class was also correct since the 3-clique
obtained had a special factorization into edge potentials. It is
generally true that marginalizing over an elimination clique takes
exponential time in the number of nodes in the clique, but the example
happened to have a special structure."
*Sep 10*: For problem 1 on the first homework set,
the following comment may be useful. Recall that a graph corresponds
to a family of probability distributions over all the nodes in
the graph. A marginal distribution is the corresponding family
of distributions over a subset of the nodes. We are asking to
find a graph that represents that family. The conditional distribution
is also a distribution over a subset of nodes (treating the values
of the conditioning nodes as constants), and again we're asking to find
a graph that represents that family of conditional distributions.
*Sep 5*: Recitation sections will be held on Mondays
from 4:30 to 6:00. They will often be held in 306 Soda Hall,
but this will change on some dates. Alternative locations
will be announced here.
*Sep 5*: Daniel Ting will hold extra office hours this
week on Friday, Sep 7, from 2 to 4 in 387 Evans Hall.
*Aug 31*: Percy Liang's office hours for the week of Sept. 3 will be
Thursday 3:30-5:30 (they are usually Weds 3-5).
*Aug 30*: The first recitation section will be held on Friday,
Aug 31, from 5:30 to 7:00 in 306 Soda Hall.