CSE 525: Randomized Algorithms and Probabilistic Analysis (WI 11)
[course description
| lectures
| assignments
| related material]
| important announcements]
important announcements
course description
Instructor: Anna Karlin,
CSE 594, tel. 543 9344
Time: Mondays and
Wednesdays in EEB 031, 12-1:20pm
Office hours: By appointment -- send email.
Teaching assistant: Punyashloka Biswal
Office hours:
- Mondays 4-5 in CSE 306.
- Tuesday 1/11, 1/18 and 2/22, in CSE 306.
- Also by appointment -- just sent email.
Course evaluation: Homework and a final exam.
About this course:
Randomization and probabilistic analysis have become fundamental tools
in modern Computer Science,
with applications ranging from combinatorial optimization to machine
learning to cryptography to
complexity theory to the design of protocols for communication
networks. Often randomized algorithms
are more efficient, and conceptually simpler and more elegant than
their deterministic counterparts.
We will cover some of the most widely used techniques for the
analysis of randomized
algorithms and the behavior of random structures from a rigorous
theoretical perspective. A (non-random)
sample of topics to be covered includes elementary examples like
fingerprinting and minimum cut, large-deviation inequalities, the probabilistic
method, martingales, random graphs,
and the analysis of Markov chains. Tools
from discrete probability will be
illustrated via their application to problems like randomized rounding,
hashing of high-dimensional data, and
load balancing in distributed systems.
related material
Required (or at least highly recommended) text:
Probability
and Computing: Randomized Algorithms and Probabilistic Analysis
by Michael Mitzenmacher and Eli Upfal.
Courses at other schools:
Suggested references:
lectures
Key (see references above):
- [MU] -- Mitzenmacher/Upfal
- [MR] -- Motwani/Raghavan
- [CG] -- Gupta/Chawla lecture notes
- [B] -- Avrim Blum lecture notes
- [HP] -- Sariel Har-Peled lecture notes
- January 3: Overview, fingerprinting ([MR] Section 7.4,
[CG] Section 2.2.1), MAX-CUT ([MU] Section 6.2.1, [CG] Section 1.4.1).
- Recommended reading: [MU] Chapters 1 and 2.
- January 5: Matrix-product verification [MU 1.3], [MR 7.1], Min-cut [MU 1.4], [CG Lecture 5].
- January 10: Coupon-collectors [MU 2.4], Markov and Chevychev Inequalities [MU 3.1-3.3], Poisson approximation (some of [MU 5.3-5.4])
- January 19: Backwards analysis -- 2D convex hull [Seidel survey, Section 4]
- January 24: Chernoff bounds and applications [MU, Chapter 4, MR 4.1]. Here are the slides.
The main application was randomized rounding analysis to approximately solve the integer multicommodity flow problem. [B Lecture 7].
- January 26: Randomized rounding of SDPs:
Goemans Williamson MAX-CUT algorithm.
- January 31: Permutation routing on the hypercube [MU Section 4.5.1, MR Section 4.2]
- February 2: Universal hashing and perfect hashing [MU Section 13.3, MR Sections 8.4-8.5]
- February 7: Markov Chains [MU Chapter 7, MR Sections 6.1-6.3, 6.6]
- February 9: Markov Chains continued.
- February 14: Random walks and electrical networks [MR, Section 6.4, B Lectures 8-9, Lecture notes
here
and here.]
- February 16: Rapidly mixing random walks and expansion [MR, Section 6.7 and these
lecture notes: Section 3 here,
Section 1 here,
first
part of these,
and these.
Additional material can be found in the rest of
lectures 8-11 in these lecture notes by Luca Trevisan. For details on an explicit construction of expanders, see
these notes.
Here is another great reference on expanders].
- February 23: Probability amplification by random walks on expanders [MR 6.8]
- February 28: Monte Carlo Methods [MU Chapter 10]
- March 2: Markov Chain Monte Carlo [MU Section 10.4,
this survey and references
therein],
coupling [MU Sections 11.1-11.2]
- March 7: Martingales [MU, Chapter 12]
- March 9: Review of what we did and didn't do.
assignments
- Homework #1 due Wednesday, January 19.
- Homework #2 due Wednesday, February 2.
- Homework #3 due Wednesday, February 16.
- Homework #4 due Wednesday, March 2.
- Homework #5 due Friday, March 11 at 1pm.