The classification game: complexity regularization through interaction

Authors:
Samarth Swarup
Affiliations:
Network Dynamics and Simulation Science Lab, Virginia Bioinformatics Institute, Virginia Polytechnic Institute and State University, Blacksburg, VA
Venue:
COIN'09 Proceedings of the 5th international conference on Coordination, organizations, institutions, and norms in agent systems
Year:
2009

Citing 12
Cited 0

Learning translation invariant recognition in massively parallel networks

Volume I: Parallel architectures on PARLE: Parallel Architectures and Languages Europe
Occam's razor

Information Processing Letters
On the necessity of Occam algorithms

STOC '90 Proceedings of the twenty-second annual ACM symposium on Theory of computing
Decision theoretic generalizations of the PAC model for neural net and other learning applications

Information and Computation
Keeping the neural networks simple by minimizing the description length of the weights

COLT '93 Proceedings of the sixth annual conference on Computational learning theory
Bagging predictors

Machine Learning
Flat minima

Neural Computation
Dynamics of multilayer networks in the vicinity of temporary minima

Neural Networks
Neural Networks: A Comprehensive Foundation

Neural Networks: A Comprehensive Foundation
Induction of Decision Trees

Machine Learning
Sharpening Occam's razor

Information Processing Letters
The classification game: combining supervised learning and language evolution

Connection Science

Quantified Score

Hi-index	0.00

Visualization

Abstract

We show that if a population of neural network agents is allowed to interact during learning, so as to arrive at a consensus solution to the learning problem, then they can implicitly achieve complexity regularization. We call this learning paradigm, the classification game. We characterize the game-theoretic equilibria of this system, and show how low-complexity equilibria get selected. The benefit of finding a low-complexity solution is better expected generalization. We demonstrate this benefit through experiments.