所属 |
工学教育研究部 工学科情報通信工学プログラム担当 |
職名 |
助教 |
外部リンク |
学位 【 表示 / 非表示 】
-
博士(工学) ( 2000年3月 東京工業大学 )
-
修士(工学) ( 1997年3月 東京大学 )
-
学士(農学) ( 1995年3月 東京大学 )
論文 【 表示 / 非表示 】
-
Computable Variants of AIXI which are More Powerful than AIXItl 査読あり
Susumu Katayama
Journal of Artificial General Intelligence 10 ( 1 ) 1 - 23 2019年4月
記述言語:英語 掲載種別:研究論文(学術雑誌)
-
Ideas for a reinforcement learning algorithm that learns programs 査読あり
Susumu Katayama
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 9782 354 - 362 2016年7月
記述言語:英語 掲載種別:研究論文(国際会議プロシーディングス) 出版者・発行元:Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
© Springer International Publishing Switzerland 2016. Conventional reinforcement learning algorithms such as Q-learning are not good at learning complicated procedures or programs because they are not designed to do that. AIXI, which is a general framework for reinforcement learning, can learn programs as the environment model, but it is not computable. AIXI has a computable and computationally tractable approximation, MC-AIXI(FAC-CTW), but it models the environment not as programs but as a trie, and still has not resolved the trade-off between exploration and exploitation within a realistic amount of computation. This paper presents our research idea for realizing an efficient reinforcement learning algorithm that retains the property of modeling the environment as programs. It also models the policy as programs and has the ability to imitate other agents in the environment. The design policy of the algorithm has two points: (1) the ability to program is indispensable for human-level intelligence, and (2) a realistic solution to the exploration/exploitation trade-off is teaching via imitation.
-
Towards Human-Level Inductive Functional Programming 査読あり
Susumu Katayama
Artificial General Intelligence, 8th International Conference, AGI 2015, LNAI 9205 111 - 120 2015年7月
記述言語:英語 掲載種別:研究論文(国際会議プロシーディングス)
-
An Analytical Inductive Functional Programming System that Avoids Unintended Programs 査読あり
Susumu Katayama
PEPM'12 Proceedings of the ACM SIGPLAN 2012 Workshop on Partial Evaluation and Program Manipulation 43 - 52 2012年1月
記述言語:英語 掲載種別:研究論文(国際会議プロシーディングス)
-
Efficient Exhaustive Generation of Functional Programs using Monte-Carlo Search with Iterative Deepening 査読あり
Susumu Katayama
PRICAI 2008: Trends in Artificial Intelligence, Lecture Notes in Artificial Intelligence, Springer Verlag 5351 199 - 211 2008年12月
記述言語:英語 掲載種別:研究論文(国際会議プロシーディングス)
書籍等出版物 【 表示 / 非表示 】
-
Trends in Functional Programming Volume 6
Edited by Marko van Eekelen( 担当: 共著 , 範囲: Chapter 8 Systematic Search for Lambda Expressions, pp. 111-126)
Intellect 2007年7月
記述言語:英語 著書種別:学術書
MISC 【 表示 / 非表示 】
-
The 8th Conference on Artificial General Intelligence, AGI-15 会議報告
片山晋
人工知能 30 ( 6 ) 2015年12月
記述言語:日本語 掲載種別:研究発表ペーパー・要旨(全国大会,その他学術会議) 出版者・発行元:人工知能学会
講演・口頭発表等 【 表示 / 非表示 】
-
BK-ADAPT: Dynamic Background Knowledge for Automating Data Transformation 国際会議
Lidia Contreras-Ochando, Cèsar Ferri, Jose Hernandez-Orallo, Fernando Martínez-Plumed, M. José Ramírez-Quintana, Susumu Katayama
The European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases
開催年月日: 2019年9月16日 - 2019年9月20日
記述言語:英語 会議種別:口頭発表(一般)
An enormous effort is usually devoted to data wrangling, the tedious process of cleaning, transforming and combining data, such that it is ready for modelling, visualisation or aggregation. Data transformation and formatting is one common task in data wrangling, which is performed by humans in two steps: (1) they recognise the specific domain of data (dates, phones, addresses, etc.) and (2) they apply conversions that are specific to that domain. However, the mechanisms to manipulate one specific domain can be unique and highly different from other domains. In this paper we present bka, a system that uses inductive programming (IP) with a dynamic background knowledge (BK) generated by a machine learning meta-model that selects the domain and/or the primitives from several descriptive features of the data wrangling problem. To show the performance of our method, we have created a web-based tool that allows users to provide a set of inputs and one or more examples of outputs, in such a way that the rest of examples are automatically transformed by the tool.
-
Automated Data Transformation with Inductive Programming and Dynamic Background Knowledge 国際会議
Lidia Contreras-Ochando, Cèsar Ferri, Jose Hernandez-Orallo, Fernando Martínez-Plumed, M. José Ramírez-Quintana, Susumu Katayama
The European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases
開催年月日: 2019年9月16日 - 2019年9月20日
記述言語:英語 会議種別:口頭発表(一般)
Data quality is essential for database integration, machine learning and data science in general. Despite the increasing number of tools for data preparation, the most tedious tasks of data wrangling -and feature manipulation in particular- still resist automation partly because the problem strongly depends on domain information. For instance, if the strings "17th of August of 2017" and "2017-08-17" are to be formatted into "08/17/2017" to be properly recognised by a data analytics tool, humans usually process this in two steps: (1) they recognise that this is about dates and (2) they apply conversions that are specific to the date domain. However, the mechanisms to manipulate dates are very different from those to manipulate addresses. This requires huge amounts of background knowledge, which usually becomes a bottleneck as the diversity of domains and formats increases. In this paper we help alleviate this problem by using inductive programming (IP) with a dynamic background knowledge (BK) fuelled by a machine learning meta-model that selects the domain, the primitives (or both) from several descriptive features of the data wrangling problem. We illustrate these new alternatives for the automation of data format transformation, which we evaluate on an integrated benchmark and code for data wrangling, which we share publicly for the community.
-
General AI Challenge に参加して 招待あり
片山晋
第9回 人工知能学会 汎用人工知能研究会(SIG-AGI) (国立情報学研究所) 人工知能学会
開催年月日: 2018年8月30日
記述言語:日本語 会議種別:口頭発表(招待・特別)
開催地:国立情報学研究所
The speaker is developing AGI agents based on the assumption that mature AGI can be achieved by the combination of ``the most general AI apart from the efficiency'' for generality and ``incremental learning'' for learning to specialize for efficiency. He applied his AGI agent to Round 1 of General AI Challenge held in 2017, and received the joint 2nd place of the qualitative prize. In this talk, he will explain how to implement the AGI agent, the devices and difficulties when applying it to the Challenge Round 1, and some thought on the General A I Challenge series.
-
MagicHaskeller-based incrementally learning agent 招待あり 国際会議
Susumu Katayama
Approaches and Applications of Inductive Programming (Schloss Dagstuhl) Leibniz Center for Informatics
開催年月日: 2017年9月17日 - 2017年9月20日
記述言語:英語 会議種別:口頭発表(一般)
開催地:Schloss Dagstuhl
-
Domain specific induction for data wrangling automation (Demo) 国際会議
Lidia Contreras-Ochando, Cèsar Ferri, José Hernández-Orallo, Fernando Martínez-Plumed, M. J. Ramírez-Quintana, Susumu Katayama
Automatic Machine Learning Workshop (Sydney, Australia)
開催年月日: 2017年8月10日
記述言語:英語 会議種別:ポスター発表
開催地:Sydney, Australia
Works(作品等) 【 表示 / 非表示 】
-
MagicHaskeller on the Web
Susumu Katayama
2012年5月23日
作品分類:ソフトウェア
-
MagicHaskeller, Analytical Synthesis Modules
Susumu Katayama
2011年4月8日
作品分類:ソフトウェア
-
MagicHaskeller, Open Source Edition
Susumu Katayama
2009年7月14日
作品分類:ソフトウェア
-
MagicHaskeller, Library Edition
Susumu Katayama
2006年5月
作品分類:ソフトウェア
-
MagicHaskeller
Susumu Katayama
2005年12月
作品分類:ソフトウェア
系統的探索に基づいた帰納関数プログラミングシステム
受賞 【 表示 / 非表示 】
-
Qualitative prize: joint 2nd place of the 1st Round of General AI Challenge
2017年10月 GoodAI
Susumu Katayama
受賞区分:国内外の国際的学術賞 受賞国:チェコ共和国
科研費(文科省・学振・厚労省)獲得実績 【 表示 / 非表示 】
-
系統的探索による帰納関数プログラミングの実用化
2009年04月 - 2012年03月
科学研究費補助金
担当区分:研究代表者
系統的網羅探索による帰納関数プログラミングアルゴリズムを開発し,実用化する
その他競争的資金獲得実績 【 表示 / 非表示 】
-
MagicHaskeller on the Web: Automated Programming as a Service
2013年09月
海外 ACM SIGPLAN PAC Grant
担当区分:研究代表者 資金種別:競争的資金
This research demonstrates a Web-based automatic programming tool, named MagicHaskeller on the Web, which can help developers programming in Ha This tool is based on the MagicHaskeller system which has been out for several years.
However, this work is not only yet another medium of its deployment, but it substantially improves the speed and offhandedness
by running it continuously as a shared service on a remote server machine.
In addition, a new user interface helping programmers to understand synthesized expressions has been devised.
The resulting system is especially helpful for beginner Haskell programmers to quickly write pure functions.
授業 【 表示 / 非表示 】
-
工学マネジメントワーク(G)
科目区分:大学院科目(修士)
-
卒業研究(情報)
科目区分:専門教育科目
-
情報工学セミナー
科目区分:専門教育科目
-
専門教育入門セミナーT(6)
科目区分:共通教育科目
-
プログラミング演習5
科目区分:専門教育科目