Leonid Kontorovich
I am a recent PhD recipient from the CMU Machine Learning department
(formerly known as CALD)
having been advised by
John
Lafferty
and
Kavita Ramanan.
Currently, I'm a postdoc at the
Weizmann Institute, being hosted by
Gideon Schechtman.
My research interests include machine learning, NLP,
automata theory,
and
concentration of measure inequalities.
I am trying to expand into functional analysis -- in particular, local Banach space theory.
Research statement, CV, and Weizmann homepage are forthcoming.
Contact
office location: WEH 7102
phone: 412 268 1120
mobile: email me and I'll give it to you
email: same as my handle (appears after the tilde in the url), at
cs.cmu.edu
TAing
10-702: Statistical Foundations of Machine Learning
15-453: Formal Languages, Automata and Computation
Current courses
Stochastic Processes (Advanced Probability II), 36-754
Regular seminars
CS Theory Lunch
Coupling
Methods Seminar
Machine Learning Lunch
If we happen to correspond and I send you LaTeX code with an undefined macro, here are the macros I use.
Link to my Hebrew University pagePapers
L. Kontorovich. Measure Concentration of Strongly Mixing Processes with Applications , PhD thesis.
L. Kontorovich. A Universal Kernel for Learning Regular Languages , to appear in MLG'07
C. Cortes, L. Kontorovich, and M. Mohri. Learning Languages with Rational Kernels. In Proceedings of The 20th Annual Conference on Learning Theory (COLT 2007). volume 4539 of Lecture Notes in Computer Science, pages 349-364, San Diego, California, June 2007. Springer, Heidelberg, Germany.
L. Kontorovich. A Linear Programming Inequality with Applications to Concentration of Measure , submitted
L. Kontorovich. Metric and Mixing Sufficient Conditions for Concentration of Measure
L. Kontorovich and K. Ramanan. Concentration Inequalities for Dependent Random Variables via the Martingale Method, submitted
L. Kontorovich. Measure Concentration of Markov Tree Processes
L. Kontorovich. Measure Concentration of Hidden Markov Processes
L. Kontorovich, J. Lafferty, D. Blei. Variational Inference and Learning for a Unified Model of Syntax, Semantics and Morphology
L. Kontorovich, C.Cortes and M. Mohri.
Learning Linearly Separable Languages,
appearing in ALT 2006
L. Kontorovich, D. Ron and Y. Singer. A Markov Model for the Acquisition of Morphological Structure. Abstract, .ps, .pdf
L. Kontorovich. Uniquely Decodable n-gram Embeddings.
Technical Report:
Abstract,
.ps,
.pdf
Expanded paper, appearing in
Theoretical Computer Science, 329/1-3 pp. 271-284, 2004.
A morphological formalism we developed in 2002
Pratt, H., Polyakov, A. and Kontorovich, L.
Evidence
for Separate Processing in the Human Brainstem of Interaural
Intensity and Temporal Disparities for Sound Lateralization.
Hearing Research, 108:1-8 (1997).
"Learning Semitic Vocalization with Hidden Markov Models" with
Daniel Lee
NIPS
version (NIPS workshop poster)
MS Word
version (much more detailed)
Talks
"Matrix Factorizations and Their Applications in Learning Theory"
PowerPoint presentation
English version
Hebrew version
Notes
1. In Learning
String-Edit Distance, Ristad and Yianilos define a
probability measure on $\Sigma^* \times \Sigma^*$ and give an
efficient method for computing it. Computing the marginal probability
requires summing over the infinitely many strings in $\Sigma^*$. We
give a simple and efficient method for
computing the marginal; I'd appreciate feedback if anyone has seen
anything similar in the literature.
[Update: on 11 May 2005, Alex Clark alerted me that he did something
similar in his
thesis]
2. My notes on Talagrand's inequality
3. An "obscure" theorem of Higman is getting automata-theoreitc
mileage
Links
People:
Mehryar Mohri
Corinna Cortes
Yoram Singer
Oren Kurland
Steve Miller
Eleazar Eskin
Cosma Shalizi
(check out Cosma's
blog)
Guy Lebanon
Daniel B. Neill
Maxim Shusteff
Alex Kontorovich
Vladimir Kontorovich
Eugene Kontorovich
David Harel
News:
הארץ
ידיעות אחרונות
Opinion Journal
Best of the Web Today
The Daily Standard
InstaPundit
LGF
Humor:
Guberman
Спору нет
http://www.theonion.com/
Михаил Жванецкий
Utilities:
andrew webmail
Quotes:
felix qui potuit rerum cognoscere causas (Virgil)
---
otium, Catulle, tibi molestum est
otio exsultas nimiumque gestis
otium et reges prius et beatas
perdidit urbes. (Catullus)
HUJI CS Dept.
http://www.cs.huji.ac.il/
Organizations:
Israel
On Campus