论文信息 - Learning temporary variable binding with dynamic links

Learning temporary variable binding with dynamic links

A novel gradient-based system for processing sequential time-varying inputs and outputs is described. With the method it is possible to train a system with time-varying inputs and outputs to use its dynamic links for temporarily binding variable contents to variable names as long as it is necessary for solving a particular task. Various learning methods for nonstationary environments are derived. Two experiments with unknown time delays illustrate the approach. A by-product of this work is the demonstration that a system consisting of two feedforward networks can solve tasks that only dynamic recurrent networks were supposed to solve.<<ETX>>

Jürgen Schmidhuber

[1] Jürgen Schmidhuber,et al. Dynamische neuronale Netze und das fundamentale raumzeitliche Lernproblem , 1990 .

[2] Ronald J. Williams,et al. Experimental Analysis of the Real-time Recurrent Learning Algorithm , 1989 .

[3] Jürgen Schmidhuber,et al. Learning to generate subgoals for action sequences , 1991, IJCNN-91-Seattle International Joint Conference on Neural Networks.

[4] P. Werbos,et al. Beyond Regression : "New Tools for Prediction and Analysis in the Behavioral Sciences , 1974 .

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

[6] T. Sejnowski,et al. Learning Algorithms for Networks with Internal and External Feedback , 1990 .

[7] Jürgen Schmidhuber,et al. Learning to generate sub-goals for action sequences , 1991 .

[8] Barak A. Pearlmutter. Learning State Space Trajectories in Recurrent Neural Networks , 1989, Neural Computation.