Increasing the Inference and Learning Speed of Tsetlin Machines with Clause Indexing

被引:4
作者
Gorji, Saeed Rahimi [1 ]
Granmo, Ole-Christoffer [1 ]
Glimsdal, Sondre [1 ]
Edwards, Jonathan [2 ]
Goodwin, Morten [1 ]
机构
[1] Univ Agder, Ctr Artificial Intelligence Res, Grimstad, Norway
[2] Temporal Comp, Newcastle Upon Tyne, Tyne & Wear, England
来源
TRENDS IN ARTIFICIAL INTELLIGENCE THEORY AND APPLICATIONS. ARTIFICIAL INTELLIGENCE PRACTICES, IEA/AIE 2020 | 2020年 / 12144卷
关键词
Tsetlin machines; Pattern recognition; Propositional logic; Frequent pattern mining; Learning automata; Pattern indexing;
D O I
10.1007/978-3-030-55789-8_60
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The Tsetlin Machine (TM) is a machine learning algorithm founded on the classical Tsetlin Automaton (TA) and game theory. It further leverages frequent pattern mining and resource allocation principles to extract common patterns in the data, rather than relying on minimizing output error, which is prone to overfitting. Unlike the intertwined nature of pattern representation in neural networks, a TM decomposes problems into self-contained patterns, represented as conjunctive clauses. The clause outputs, in turn, are combined into a classification decision through summation and thresholding, akin to a logistic regression function, however, with binary weights and a unit step output function. In this paper, we exploit this hierarchical structure by introducing a novel algorithm that avoids evaluating the clauses exhaustively. Instead we use a simple look-up table that indexes the clauses on the features that falsify them. In this manner, we can quickly evaluate a large number of clauses through falsification, simply by iterating through the features and using the look-up table to eliminate those clauses that are falsified. The look-up table is further structured so that it facilitates constant time updating, thus supporting use also during learning. We report up to 15 times faster classification and three times faster learning on MNIST and Fashion-MNIST image classification, and IMDb sentiment analysis.
引用
收藏
页码:695 / 708
页数:14
相关论文
共 15 条
[1]   The regression Tsetlin machine: a novel approach to interpretable nonlinear regression [J].
Abeyrathna, K. Darshana ;
Granmo, Ole-Christoffer ;
Zhang, Xuan ;
Jiao, Lei ;
Goodwin, Morten .
PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY A-MATHEMATICAL PHYSICAL AND ENGINEERING SCIENCES, 2020, 378 (2164)
[2]  
[Anonymous], 1989, Learning Automata: An Introduction
[3]   Using the Tsetlin Machine to Learn Human-Interpretable Rules for High-Accuracy Text Categorization With Medical Applications [J].
Berge, Geir Thore ;
Granmo, Ole-Christoffer ;
Tveit, Tor Oddbjorn ;
Goodwin, Morten ;
Jiao, Lei ;
Matheussen, Bernt Viggo .
IEEE ACCESS, 2019, 7 :115134-115146
[4]  
GITTINS JC, 1979, J ROY STAT SOC B MET, V41, P148
[5]   A Tsetlin Machine with Multigranular Clauses [J].
Gorji, Saeed Rahimi ;
Granmo, Ole-Christoffer ;
Phoulady, Adrian ;
Goodwin, Morten .
ARTIFICIAL INTELLIGENCE XXXVI, 2019, 11927 :146-151
[6]   Learning automata-based solutions to the nonlinear fractional knapsack problem with applications to optimal resource allocation [J].
Granmo, Ole-Christoffer ;
Oommen, B. John ;
Myrer, Svein Arild ;
Olsen, Morten Goodwin .
IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART B-CYBERNETICS, 2007, 37 (01) :166-175
[7]  
Granmo OC, 2021, Arxiv, DOI arXiv:1804.01508
[8]  
Granmo OC, 2019, Arxiv, DOI [arXiv:1905.09688, DOI 10.48550/ARXIV.1905.09688, 10.48550/arXiv.1905.09688]
[9]  
Haugland V, 2014, LECT NOTES COMPUT SC, V8615, P1, DOI 10.1007/978-3-662-44509-9_1
[10]  
Phoulady A., 2020, 9 INT WORKSH STAT RE