Generalized compressed network search

Authors:
Rupesh Kumar Srivastava;Jürgen Schmidhuber;Faustino Gomez
Affiliations:
USI-SUPSI, Manno-Lugano, Switzerland;USI-SUPSI, Manno-Lugano, Switzerland;USI-SUPSI, Manno-Lugano, Switzerland
Venue:
Proceedings of the 14th annual conference companion on Genetic and evolutionary computation
Year:
2012

Citing 2
Cited 0

Discovering neural nets with low Kolmogorov complexity and high generalization capability

Neural Networks
Evolving neural networks in compressed weight space

Proceedings of the 12th annual conference on Genetic and evolutionary computation

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents initial results of Generalized Compressed Network Search (GCNS), a method for automatically identifying the important frequencies for neural networks encoded as a set of Fourier-type coefficients (i.e. "compressed" networks). GCNS achieves better compression than our previous approach, and promises better generalization capabilities. Results for a high-dimensional Octopus arm control problem show that a high fitness 3680-weight network can be encoded using less than 10 coefficients, using the frequencies identified by GCNS.