Watanabe, Kazuho

Japanese

Affiliation	Department of Computer Science and Engineering
Title	Associate Professor
Fields of Research	Statistical Learning Theory / Machine Learning
Degree	Ph. D. (Tokyo Institute of Technology)
Academic Societies	IEICE / IEEE
E-mail	wkazuho@cs Please append ".tut.ac.jp" to the end of the address above.
Laboratory website URL	http://www.lisl.cs.tut.ac.jp
Researcher information URL（researchmap）	Researcher information

Research

Machine learning techniques are widely used for pattern recognition, robot control, and so on. We study fundamental theories of machine learning based on statistical and information theoretic methods, and apply them to data analysis problems.

Theme1：Analysis and Development of Statitstical Learning Methods

Overview

Machine learning techniques are widely used for various applications such as pattern recognition and robot control. We aim to analyze and develop learning and inference methods, and apply them to problems such as data analysis and visualization.

Selected publications and works

Omae, A. and Watanabe, K. (2022), "Approximate empirical Bayes estimation of the regularization parameter in l1 trend filtering, Proc. of ISIT2022, 462-467.
Yoshida, T. and Watanabe, K. (2018), Empirical Bayes estimation for L_1 regularization: a detailed analysis in the one-parameter lasso model, IEICE Transactions on Fundamentals, E101-A(12), 2184-2191.
Kobayashi, M. and Watanabe, K. (2017), A rate-distortion theoretic view of Dirichlet process means clustering, IEICE Transactions on Fundamentals, J100-A(12), 475-486 (in Japanese).
Watanabe, K. and Roos, T. (2015), Achievability of asymptotic minimax regret by horizon-dependent and horizon-independent strategies, Journal of Machine Learning Research, 16, 2357-2375.
Watanabe, K. and Ikeda, S. (2015), Entropic risk minimization for nonparametric estimation of mixing distributions, Machine Learning, 99(1), 119-136.

Keywords

Bayesian inference, learning algorithm, data visualization

Theme2：Rate-Distortion Theory

Overview

Rate-distortion function

Rate-distortion functions show the minimum information content required for reconstructing compressed data under allowed distortion levels.
We aim at evaluating rate-distortion functions of distortion measures used in practical learning algorithms and information sources modeling real data generation processes.

Selected publications and works

Konabe, R. and Watanabe, K. (2018), Sparse regression code with sparse dictionary for absolute error criterion, Proc. of ISIT2018, 1515-1519.
Watanabe, K. (2017), Projection to mixture families and rate-distortion bounds with power distortion measures, Entropy, Special Issue: Information Geometry II, 19(6), 262.
Watanabe, K. and Ikeda, S. (2016), Rate-distortion functions for gamma-type sources under absolute-log distortion measure, IEEE Transactions on Information Theory, 62(10), pp.5496-5502.
Watanabe, K. (2015), Vector quantization based on epsilon-insensitive mixture models, Neurocomputing, 165, 32-37.

Keywords

rate-distortion function, lossy data compression

Title of class

Information Theory and Coding: B12630030(Dept. EEIE), B13630010(Dept. CSE)
Advanced Statistical Machine Learning: M23620080

Others (Awards, Committees, Board members)

Awards:
Encouragement Award (IEICE SITA Subsociety), 2013.
Best Paper Method Award (German Classification Society), 2010.
Best Paper Award (Japanese Neural Network Society), 2008.