|Kohei Hatano||Last modified date：2018.06.15|
Country of degree conferring institution (Overseas)
Field of Specialization
ORCID(Open Researcher and Contributor ID)
Total Priod of education and research career in the foreign country
My main research interest includes Machine Learning, in particular, design and analyses of robust and efficient machine learning algorithms from a theoretical perspective. I am also affiliated with Department of Informatics and the Learning Analytics Center for analyzing educational data.
Research InterestsMembership in Academic Society
- Open Science
keyword : open access, open data, open science
- Design and analysis of online prediction algorithms
keyword : online learning, computational learning theory, machine learning, optimization, ranking
- Analysis and depelopment of boosting algorithms
keyword : boosting, machine learning, computational learning theory, data mining
2000.04[Theoretical analyses of boosting methods] Boosting is a technique for constructing a highly accurate classifier by combining many ``weak'' classifiers. Although boosting has become a fundamental tool in machine learning and data mining these days, its theoretical propeties are yet to be understood. We have analysed theoretical properties of boosting in order to design more efficient boosting methods. As a result, we clarified a property which might be a key for further improvement. So far, we are developping a new boosting algorithm based on our analysis..
|1.||Boosting - methods for design of learning algorithms -.|
|2.||Report on The 21st Annual Conference on Learning Theory (COLT 2008) .|
|3.||Frontier of Theoretical Computer Science.|
|2.||Takahiro Fujita, Kohei Hatano, Shuji Kijima, Eiji Takimoto, Online Linear Optimization over Permutations with Precedence Constraints
, NIPS 2014 Workshop on Discrete Optimization in Machine Learning(DISCML), 2014.12.
|3.||Issei Matsumoto, Kohei Hatano, Eiji Takimoto, Online Prediction with Bradley-Terry Models, NIPS 2014 Workshop on Analysis of Rank Data: Confluence of Social Choice, Operations Research, and Machine Learning, 2014.12.|
|4.||畑埜 晃平, Combinatorial Online Prediction via Metarounding, TCE Guest Lecture, 2013.12.|
|5.||Takahiro Fujita, Kohei Hatano, Eiji Takimoto, Combinatorial Online Prediction Using Offline Approximation Algorithms, The sixth Annual Meeting of Asian Association for Algorithms and Computation (AAAC2013), 2013.04.|
|6.||Daiki Suehiro, Kohei Hatano, Shuji Kijima, Eiji Takimoto, Kiyohito Nagano, Online Prediction over Base Polyhedra
, NIPS 2012 Workshop on Discrete Optimization in Machine Learning(DISCML), 2012.12.
|7.||Online Rank Aggregation.|
|8.||Online Rank Aggregation.|
|9.||Learning evaluation functions for Shogi via bipertite ranking learning with SVMs .|
|10.||Learning evaluation functions for Shogi via bipertite ranking learning with SVMs .|
|11.||Online Rank Aggregation.|
|12.||Online Rank Aggregation.|
|13.||Online Learning Based On Maximum Entropy Principle.|
|14.||Online Learning of Maximum p-Norm Margin Classifiers with Bias.|
|15.||Efficient online learning of linear threshold functions with large biases..|
|16.||Boosting using classifiers with nearly one-sided error.|
- We show that one of machine learning techniques called “boosting” can be efficiently computed over compressed data using non-determinist Zero-Suppressed Binary Decision Diagrams (ZDDs). A big advantage of this scheme is the ability to perform analysis using big data with minimal space. An interesting point in this research is that the combinatorial online prediction, a seemingly unrelated machine learning concept is successfully adopted for utilizing compressed data.
- JSAI Incentive Award
- JSAI Incentive Award