Genetic Algorithms for Multi-classifier System Configuration: A Case Study in Character Recognition

Abstract

We describe a multiple classifier system which incorporates an automatic self-configuration scheme based on genetic algorithms. Our main interest in this paper is focused on exploring the statistical properties of the resulting multi-expert configurations. To this end we initially test the proposed system on a series of tasks of increasing difficulty drawn from the domain of character recognition. We then proceed to investigate the performance of our system not only in comparison to that of its constituent classifiers, but also in comparison to an independent set of individually optimised classifiers. Our results illustrate that significant gains can be obtained by integrating a genetic algorithm based optimisation process into multi-classifier schemes both in the performance enhancement and in the reduction of its volatility, especially as the task domain becomes more complex.

The Halting Probability in Von Neumann Architectures

“…An interesting alternative is to pretrain the experts and optimise the combination rule. With a small number of experts [Sirlantzis et al, 2001] and a simple voting rule it might be feasible to try all possible combinations of experts. However there are 2 n (where n = number of experts) possible combinations in such a voting scheme.…”

Optimization of Weights in a Multiple Classifier Handwritten Word Recognition System Using a Genetic Algorithm

“…the application was the recognition of handwritten digits and the combined classifiers were not created by an ensemble creation method, but were each separately designed by hand. In [30] a genetic algorithm was used for the selection of a subset of classifiers from an ensemble, which is equivalent to weight optimization using only the weights 0 and 1. Another application of a genetic algorithm in a multiple classifier framework has been proposed in [16].…”

Recognition of cursive Roman handwriting: past, present and future

“…Particularly in handwriting recognition the use of multiple classifier systems has been advocated by many authors. Examples include [31,33,88,108,114].…”

“…More sophisticated combination procedures use the score values output by the individual classifiers as input for a trainable classifier, for example a neural network that acts as a combiner [33]. Another interesting approach is to view the selection of the individual classifiers of the ensemble, including their weights and perhaps even the combination procedure, as an optimization problem and find the solution by means of evolutionary search procedures [108].…”