Shai Fine

The Institute of Computer Science
The Hebrew University of Jerusalem
Jerusalem 91904

Lab : +972-2-6585775
Home : +972-9-7447093

Research Interest:

I am interested in machine learning problems that arise in human-machine interaction and involve temporal modeling and statistical inference. Such are the modeling of text, NLP and speech recognition. The methods which I'm using in my research emerge from combinations of various research fields such as Learning Theory, Applicable Unsupervised Learning, Hierarchical Modeling and Information Theory.

Publications list:

    The Hierarchical Hidden Markov Model: Analysis and Applications

    Shai Fine , Yoram Singer And Naftali Tishby
    Prepared: March 1995, Appeared: Machine Learning, Vol 32, pp 41-62, 1998.
    23 pages 366303 bytes.

    Agnostic Classification of Markovian Sequences

    Ran El-Yaniv, Shai Fine And Naftali Tishby
    Prepared: June 1997, revised January 1998.
    Appeared: Advances in Neural Information Processing Systems 10, Ed. M. Jordan, M. Kearns and A. Solla, Vol 10, pp 465-71, 1997.
    7 pages 360,847 bytes.

    Query by Committee, Linear Separation and Random Walks

    Ran Bachrach, Shai Fine and Eli Shamir
    Prepared: September 1998. Accepted to EuroColt-99. 16 pages.
    Journal version: TCS 284(1):25-51(2002)

    Noise Tolerant Learning using Early Predictors

    Shai Fine, Ran Gilad-Bachrach, Eli Shamir and Naftali Tishby
    Prepared: May 1999. 7 pages.

    Noise Tolerant Learnability via the Dual Learning Problem

    Shai Fine, Ran Gilad-Bachrach, Shahar Mendelson and Naftali Tishby
    Prepared: June 1999. Accepted: NCST-99. 12 pages.

    Efficient SVM Training Using Low-Rank Kernel Representation

    Shai Fine and Katya Scheinberg
    Presented: NIPS workshop on "New Perspectives in Kernel-Based Learning Methods", 2000.
    Journal version: JMLR (Special Issue on Kernel Methods) 2(Dec):243-264, 2001.

    Incremental Learning and Selective Sampling via Parametric Optimization Framework for SVM

    Shai Fine and Katya Scheinberg
    Presented: Neural Information Processing Systems (NIPS), 2001.

    A Hybrid GMM/SVM Approach to Speaker Identification

    Shai Fine, Jiri Navratil and Ramesh Gopinath
    Presented: The International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2001.

    Enhancing GMM Scores Using SVM 'Hints'

    Shai Fine, Jiri Navratil and Ramesh Gopinath
    Presented: The 7th European Conference on Speech Communication and Technology (Eurospeech), 2001.

    Digit Recognition in Noisy Environments via a Sequential GMM/SVM System

    Shai Fine, George Saon and Ramesh Gopinath
    Presented: The International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2002.

    Discriminative Feature Selection via Multiclass Variable Memory Markov Model

    Noam Slonim, Gill Bejerano, Shai Fine and Naftali Tishby
    Presented: The 19th International Conference on Machine Learning (ICML), 2002.
    Journal Version: EURASIP Journal on Applied Signal Processing (JASP), Special issue on Unstructured Information Management from Multimedia Data Sources, 2003.

    INCAS: An Incremental Active Set Method for SVM

    Shai Fine and Katya Scheinberg
    Presented: The 6th Kernel Machines Workshop "On Learning Kernels", (NIPS), 2002.

    Coverage directed test generation for functional verification using Bayesian networks

    Shai Fine and Avi Ziv
    Presented: The 40th Design Automation Conference (DAC), 2003.

    Enhancing the Control and Efficiency of the Covering Process

    Shai Fine and Avi Ziv
    Accepted: IEEE International High Level Design Validation and Test Workshop (HLDVT), 2003.

Technical Reports:

    Stochastic Modeling and Recognition of Solo Instruments using Harmonic and Multi Band Noise Features

    Shlomo Dubnov and Shai Fine
    Prepared: February 1999.


Back to CS HUJI Home Page