Abstract
Littlewood and Miller [4] present a statistical framework for dealing with coincident failures in multiversion software systems. They develop a theoretical model that holds the promise of high system reliability through the use of multiple, diverse sets of alternative versions. In this paper, we adapt their framework to investigate the feasibility of exploiting the diversity observable in multiple populations of neural networks developed using diverse methodologies. We evaluate the generalisation improvements achieved by a range of methodologically diverse network generation processes. We attempt to order the constituent methodological features with respect to their potential for use in the engineering of useful diversity. We also define and explore the use of relative measures of the diversity between version sets as a guide to the potential for exploiting interset diversity.
Similar content being viewed by others
References
Partridge D, Griffith N. Strategies for improving neural network generalisation. Neural Computing and Applications 1995; 3: 27–37
Littlewood B, Miller DR. Conceptual modeling of coincident failures in multiversion software. IEEE Transactions on Software Engineering 1989; 15(12): 1596–1614
Hornik AK, Stinchcombe AM, White H. Multilayer feedforward networks are universal approximators. Neural Networks 1989; 2(5): 359–366
Rumelhart DE, Hinton GE, Williams RJ. Learning internal representation by error propagation. In DE Rumelhart, JL McClelland, eds, Parallel Distributed Processing. Explorations in the Microstructure of Cognition. Volume 1: Foundations, 318–362. MIT Press, Cambridge, MA, 1986
Moody gnJ, Darken C. Learning with localized receptive fields. In D Touretzky, GE Hinton, TJ Sejnowski, eds, Proc Connectionist Models Summer School, 1988, 133–143, Morgan Kaufman
Holden SB, Rayner PJW. Generalization and PAC learning: Some new results for the class of generalized single layer networks. IEEE Transactions on Neural Networks 1993 (submitted)
Partridge D, Yates WB. Replicability of neural computing experiments. Research Report R305, University of Exeter, 1994
Widrow B, Hoff ME. Adaptive switching circuits. In JA Anderson, E Rosenfeld, eds, Neurocomputing: Foundations of Research, 126–134. MIT Press, Cambridge, MA, 1960
Knight JC, Leveson NG. An experimental evaluation of the assumption of independence in multiversion programming. IEEE Transactions on Software Engineering 1986; 12(1): 96–109
Partridge D. Network generalization differences quantified. Neural Networks (in press)
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Yates, W.B., Partridge, D. Use of methodological diversity to improve neural network generalisation. Neural Comput & Applic 4, 114–128 (1996). https://doi.org/10.1007/BF01413747
Issue Date:
DOI: https://doi.org/10.1007/BF01413747