Weighted Automata Extraction from Recurrent Neural Networks via Regression on State Spaces
Weighted Automata Extraction from Recurrent Neural Networks via Regression on State Spaces
复制标题
DOI：
10.1609/aaai.v34i04.5977
复制
发表时间：
2019-04
期刊：
影响因子：
--
通讯作者：
Takamasa Okudono;Masaki Waga;Taro Sekiyama;I. Hasuo
中科院分区：
文献类型：
--
作者：
Takamasa Okudono;Masaki Waga;Taro Sekiyama;I. Hasuo
我们提出了一种从复发性神经网络（RNN）提取加权有限自动机（WFA）的方法。我们的方法基于Balle和Mohri的WFA学习算法，而Balle和Mohri则是Angluin经典L*算法的扩展。我们的技术新颖性是将回归方法用于所谓的等价查询，从而利用RNN的内部状态空间来优先考虑反例候选者。这样，我们实现了提取DFA的Weiss，Goldberg和Yahav最近工作的定量/加权扩展。我们通过实验评估提取的WFA的准确性，表现性和效率。
We present a method to extract a weighted finite automaton (WFA) from a recurrent neural network (RNN). Our method is based on the WFA learning algorithm by Balle and Mohri, which is in turn an extension of Angluin's classic L* algorithm. Our technical novelty is in the use of regression methods for the so-called equivalence queries, thus exploiting the internal state space of an RNN to prioritize counterexample candidates. This way we achieve a quantitative/weighted extension of the recent work by Weiss, Goldberg and Yahav that extracts DFAs. We experimentally evaluate the accuracy, expressivity and efficiency of the extracted WFAs.