Title: Interpretation of Recurrent Neural Networks

Authors: Morten With Pedersen and Jan Larsen
         CONNECT, Department of Mathematical Modelling, Building 321
         Technical University of Denmark
         DK-2800 Lyngby, Denmark
         Phones: + 45 4525 + ext. 3920,3923
         Fax: + 45 45872599
         emails: mwp@imm.dtu.dk, jl@imm.dtu.dk

Abstract:
This paper addresses techniques for interpretation and characterization
of trained recurrent nets for time series problems. In particular, we
focus on assessment of effect memory and suggest an operational
definition of memory. Further we discuss the evaluation of learning
curves. Various numerical experiments on time series prediction problems 
are used to illustrate the potential of the suggested method.



In proccedings of the IEEE Workshop on Neural Networks for Signal
Processing VII, IEEE Press, 1997. NNSP'97 is taking palce in  
Florida, September, 1997.