KATAYAMA Susumu

写真a

Affiliation

Engineering educational research section Information and Communication Technology Program

Title

Assistant Professor

External Link

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

     More details

    Language:English   Publishing type:Research paper (scientific journal)  

    DOI: 10.2478/jagi-2019-0001

  • 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

     More details

    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.

    DOI: 10.1007/978-3-319-41649-6_36

    Scopus

  • Towards Human-Level Inductive Functional Programming Reviewed

    Susumu Katayama

    Artificial General Intelligence, 8th International Conference, AGI 2015, LNAI 9205   111 - 120   2015.7

     More details

    Language:English   Publishing type:Research paper (international conference proceedings)  

    DOI: 10.1007/978-3-319-21365-1_12

  • 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

     More details

    Language:English   Publishing type:Research paper (international conference proceedings)  

    DOI: 10.1145/2103746.2103758

  • 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

     More details

    Language:English   Publishing type:Research paper (international conference proceedings)  

display all >>

Books 【 display / non-display

  • Trends in Functional Programming Volume 6

    ( Role: Joint author)

    2007.7 

     More details

    Language:English Book type:Scholarly book

MISC 【 display / non-display

  • Report on AGI-15

    Susumu Katayama

    30 ( 6 )   2015.12

     More details

    Language:Japanese   Publishing type:Research paper, summary (national, other academic conference)   Publisher:Japanese Society for Artificial Intelligence  

Presentations 【 display / non-display

  • 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 

     More details

    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 

     More details

    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

     More details

    Event date: 2018.8.30

    Language:Japanese   Presentation type:Oral presentation (invited, special)  

    Venue:National Institute for Informatics  

  • MagicHaskeller-based incrementally learning agent Invited International conference

    Susumu Katayama

    Approaches and Applications of Inductive Programming  (Schloss Dagstuhl)  Leibniz Center for Informatics

     More details

    Event date: 2017.9.17 - 2017.9.20

    Language:English   Presentation type:Oral presentation (general)  

    Venue:Schloss Dagstuhl  

  • Domain specific induction for data wrangling automation (Demo) International conference

    Lidia Contreras-Ochando, Cèsar Ferri, José Hernández-Orallo, Fernando Martínez-Plumed, M. J. Ramírez-Quintana, Susumu Katayama

    Automatic Machine Learning Workshop  (Sydney, Australia) 

     More details

    Event date: 2017.8.10

    Language:English   Presentation type:Poster presentation  

    Venue:Sydney, Australia  

display all >>

Works 【 display / non-display

  • MagicHaskeller on the Web

    Susumu Katayama

    2012.5.23

     More details

    Work type:Software  

  • MagicHaskeller, Analytical Synthesis Modules

    Susumu Katayama

    2011.4.8

     More details

    Work type:Software  

  • MagicHaskeller, Open Source Edition

    Susumu Katayama

    2009.7.14

     More details

    Work type:Software  

  • MagicHaskeller, Library Edition

    Susumu Katayama

    2006.5

     More details

    Work type:Software  

  • MagicHaskeller

    Susumu Katayama

    2005.12

     More details

    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

     More details

    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

    2009.04 - 2012.03

    科学研究費補助金 

      More details

    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