|
Affiliation |
Engineering educational research section Information and Communication Technology Program |
|
Title |
Assistant Professor |
|
External Link |
|
KATAYAMA Susumu
|
|
|
Degree 【 display / non-display 】
-
Ph.D. in Engineering ( 2000.3 Tokyo Institute of Technology )
-
Master of Engineering ( 1997.3 The University of Tokyo )
-
Bachelor in Agriculture ( 1995.3 The University of Tokyo )
Research Areas 【 display / non-display 】
-
Informatics / Intelligent informatics
-
Informatics / Software
-
Informatics / Theory of informatics
Papers 【 display / non-display 】
-
Computable Variants of AIXI which are More Powerful than AIXItl Reviewed
Susumu Katayama
Journal of Artificial General Intelligence 10 ( 1 ) 1 - 23 2019.4
Language：English Publishing type：Research paper (scientific journal)
-
Ideas for a reinforcement learning algorithm that learns programs Reviewed
Susumu Katayama
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 9782 354 - 362 2016.7
Language：English Publishing type：Research paper (international conference proceedings) Publisher：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 Reviewed
Susumu Katayama
Artificial General Intelligence, 8th International Conference, AGI 2015, LNAI 9205 111 - 120 2015.7
Language：English Publishing type：Research paper (international conference proceedings)
-
An Analytical Inductive Functional Programming System that Avoids Unintended Programs Reviewed
Susumu Katayama
PEPM'12 Proceedings of the ACM SIGPLAN 2012 Workshop on Partial Evaluation and Program Manipulation 43 - 52 2012.1
Language：English Publishing type：Research paper (international conference proceedings)
-
Efficient Exhaustive Generation of Functional Programs using Monte-Carlo Search with Iterative Deepening Reviewed
Susumu Katayama
PRICAI 2008: Trends in Artificial Intelligence, Lecture Notes in Artificial Intelligence, Springer Verlag 5351 199 - 211 2008.12
Language：English Publishing type：Research paper (international conference proceedings)
Books 【 display / non-display 】
-
Trends in Functional Programming Volume 6
（ Role： Joint author）
2007.7
Language：English Book type：Scholarly book
MISC 【 display / non-display 】
-
Report on AGI-15
Susumu Katayama
30 ( 6 ) 2015.12
Language：Japanese Publishing type：Research paper, summary (national, other academic conference) Publisher：Japanese Society for Artificial Intelligence
Presentations 【 display / non-display 】
-
Improving n-step return in reinforcement learning
2024.3.14
Event date： 2024.3.13 - 2024.3.14
Language：Japanese Presentation type：Oral presentation (general)
-
Plug-in development to propose functions using Haskell Language Server extension of VSCode
Record of Joint Conference of Electrical and Electronics Engineers in Kyushu 2023.9 Committee of Joint Conference of Electrical, Electronics and Information Engineers in Kyushu
Event date： 2023.9.7 - 2023.9.8
Language：Japanese Presentation type：Oral presentation (general)
-
BK-ADAPT: Dynamic Background Knowledge for Automating Data Transformation International conference
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
Event date： 2019.9.16 - 2019.9.20
Language：English Presentation type：Oral presentation (general)
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 International conference
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
Event date： 2019.9.16 - 2019.9.20
Language：English Presentation type：Oral presentation (general)
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.
-
Experience on General AI Challenge Invited
Susumu Katayama
Japanese Society for Artificial Intelligence
Event date： 2018.8.30
Language：Japanese Presentation type：Oral presentation (invited, special)
Venue：National Institute for Informatics
Works 【 display / non-display 】
-
MagicHaskeller on the Web
Susumu Katayama
2012.5.23
Work type：Software
-
MagicHaskeller, Analytical Synthesis Modules
Susumu Katayama
2011.4.8
Work type：Software
-
MagicHaskeller, Open Source Edition
Susumu Katayama
2009.7.14
Work type：Software
-
MagicHaskeller, Library Edition
Susumu Katayama
2006.5
Work type：Software
-
MagicHaskeller
Susumu Katayama
2005.12
Work type：Software
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
Awards 【 display / non-display 】
-
Qualitative prize: joint 2nd place of the 1st Round of General AI Challenge
2017.10 GoodAI
Susumu Katayama
Award type：International academic award (Japan or overseas) Country：Czech Republic
Grant-in-Aid for Scientific Research 【 display / non-display 】
-
系統的探索による帰納関数プログラミングの実用化
Grant number：21650032 2009.04 - 2012.03
科学研究費補助金 挑戦的萌芽研究
Authorship：Principal investigator
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