LEARNING THE INITIAL-STATE OF A 2ND-ORDER RECURRENT NEURAL-NETWORK DURING REGULAR-LANGUAGE INFERENCE

被引:24
作者
FORCADA, ML
CARRASCO, RC
机构
关键词
D O I
10.1162/neco.1995.7.5.923
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Recent work has shown that second-order recurrent neural networks (2ORNNs) may be used to infer regular languages. This paper presents a modified version of the real-time recurrent learning (RTRL) algorithm used to train 2ORNNs, that learns the initial state in addition to the weights. The results of this modification, which adds extra flexibility at a negligible cost in time complexity, suggest that it may be used to improve the learning of regular languages when the size of the network is small.
引用
收藏
页码:923 / 930
页数:8
相关论文
共 7 条
  • [1] GILES CL, 1992, ADV NEUR IN, V4, P317
  • [2] LEARNING AND EXTRACTING FINITE STATE AUTOMATA WITH 2ND-ORDER RECURRENT NEURAL NETWORKS
    GILES, CL
    MILLER, CB
    CHEN, D
    CHEN, HH
    SUN, GZ
    LEE, YC
    [J]. NEURAL COMPUTATION, 1992, 4 (03) : 393 - 405
  • [3] SIEGELMANN HT, 1992, INFORMATION PROCESSI, V1, P329
  • [4] TOMITA M, 1982, 4TH P ANN COGN SCI C, P105
  • [5] INDUCTION OF FINITE-STATE LANGUAGES USING 2ND-ORDER RECURRENT NETWORKS
    WATROUS, RL
    KUHN, GM
    [J]. NEURAL COMPUTATION, 1992, 4 (03) : 406 - 414
  • [6] WATROUS RL, 1992, ADV NEURAL INFORMATI, V4, P306
  • [7] A Learning Algorithm for Continually Running Fully Recurrent Neural Networks
    Williams, Ronald J.
    Zipser, David
    [J]. NEURAL COMPUTATION, 1989, 1 (02) : 270 - 280