Learning to Generate Combinatorial Action Sequences Utilizing the Initial Sensitivity of Deterministic Dynamical Systems

This study shows how sensory-action sequences of imitating finite state machines (FSMs) can be learned by utilizing the deterministic dynamics of recurrent neural networks (RNNs). Our experiments indicated that each possible combinatorial sequence can be recalled by specifying its respective initial state value and also that fractal structures appear in this initial state mapping after the learning converges. We also observed that the sequences of mimicking FSMs are encoded utilizing the transient regions rather than the invariant sets of the evolved dynamical systems of the RNNs.

[1]  J. Pollack.  The Induction of Dynamical Recognizers , 1996, Machine Learning.

[2]  Richard S. Sutton,et al.  A Menu of Designs for Reinforcement Learning Over Time , 1995 .

[3]  John F. Kolen,et al.  Exploring the computational capabilities of recurrent neural networks , 1995 .

[4]  Holger H. Hoos,et al.  On the Run-time Behaviour of Stochastic Local Search Algorithms for SAT , 1999, AAAI/IAAI.

[5]  J. McCarthy.  Situations, Actions, and Causal Laws , 1963 .

[6]  D M Wolpert,et al.  Multiple paired forward and inverse models for motor control , 1998, Neural Networks.

[7]  Panos M. Pardalos,et al.  Approximate solution of weighted MAX-SAT problems using GRASP , 1996, Satisfiability Problem: Theory and Applications.

[8]  Young,et al.  Inferring statistical complexity. , 1989, Physical review letters.

[9]  Geoffrey E. Hinton,et al.  Learning internal representations by error propagation , 1986 .

[10]  James L. McClelland,et al.  Parallel distributed processing: explorations in the microstructure of cognition, vol. 1: foundations , 1986 .

[11]  Hartmut Schmeck,et al.  Information Exchange in Multi Colony Ant Algorithms , 2000, IPDPS Workshops.

[12]  Jun Tanji,et al.  Role for supplementary motor area cells in planning several movements ahead , 1994, Nature.

[13]  Marco Dorigo,et al.  Distributed Optimization by Ant Colonies , 1992 .

[14]  M. Kawato,et al.  Formation and control of optimal trajectory in human multijoint arm movement , 1989, Biological Cybernetics.

[15]  S. Rossi,et al.  Effects of repetitive transcranial magnetic stimulation on movement-related cortical activity in humans. , 2000, Cerebral cortex.

[16]  Jeffrey L. Elman,et al.  Finding Structure in Time , 1990, Cogn. Sci..

[17]  Michael I. Jordan.  Attractor dynamics and parallelism in a connectionist sequential machine , 1990 .

[18]  I. Tsuda.  Toward an interpretation of dynamic neural activity in terms of chaotic dynamical systems. , 2001, The Behavioral and brain sciences.

[19]  Jun Tani,et al.  Model-based learning for mobile robot navigation from the dynamical systems perspective , 1996, IEEE Trans. Syst. Man Cybern. Part B.

[20]  Panos M. Pardalos,et al.  A Parallel GRASP for MAX-SAT Problems , 1996, PARA.

[21]  Richard F. Hartl,et al.  Cooperative Ant Colonies for Optimizing Resource Allocation in Transportation , 2001, EvoWorkshops.

[22]  Habiba Drias,et al.  Scatter Search with Random Walk Strategy for SAT and MAX-W-SAT Problems , 2001, IEA/AIE.

[23]  Marc Gravel,et al.  PARALLEL IMPLEMENTATION OF AN ANT COLONY OPTIMIZATION METAHEURISTIC WITH OPENMP , 2001 .

[24]  C. Lee Giles,et al.  Higher Order Recurrent Networks and Grammatical Inference , 1989, NIPS.

[25]  Michael I. Jordan,et al.  Forward Models: Supervised Learning with a Distal Teacher , 1992, Cogn. Sci..

[26]  Naohiro Fukumura,et al.  Embedding a grammatical description in deterministic chaos: an experiment in recurrent neural learning , 1995, Biological Cybernetics.