Extracting Automata from Recurrent Neural Networks Using Queries and Counterexamples

被引：0

作者：

Weiss, Gail ^{[1
]}

Goldberg, Yoav ^{[2
]}

Yahav, Eran ^{[1
]}

机构：

[1] Technion, Haifa, Israel

[2] Bar Ilan Univ, Ramat Gan, Israel

来源：

INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 80 | 2018年 / 80卷

基金：

欧洲研究理事会;

关键词：

D O I：

暂无

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

We present a novel algorithm that uses exact learning and abstraction to extract a deterministic finite automaton describing the state dynamics of a given trained RNN. We do this using Angluin's L* algorithm as a learner and the trained RNN as an oracle. Our technique efficiently extracts accurate automata from trained RNNs, even when the state vectors are large and require fine differentiation.

引用

页数：10

共 20 条

[1] LEARNING REGULAR SETS FROM QUERIES AND COUNTEREXAMPLES [J].

ANGLUIN, D .

INFORMATION AND COMPUTATION, 1987, 75 (02) :87-106

[2]

[Anonymous], 2017, CoRR

[3]

[Anonymous], 1997, Neural Computation

[4]

Antiga L., 2017, P NIPS

[5]

Boser B. E., 1992, Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory, P144, DOI 10.1145/130385.130401

[6]

Cechin AL, 2003, SCCC 2003: XXIII INTERNATIONAL CONFERENCE OF THE CHILEAN COMPUTER SCIENCE SOCIETY, PROCEEDINGS, P73

[7]

Cho K., 2014, ARXIV, DOI 10.3115/v1/w14-4012

[8]

Chung J., 2014, ARXIV

[9] FINDING STRUCTURE IN TIME [J].

ELMAN, JL .

COGNITIVE SCIENCE, 1990, 14 (02) :179-211

[10]

Giles C. L., 1990, ADV NEURAL INFORM PR, p380{387

← 1 2 →