A neural evolutionary approach to financial modeling

Abstract

This paper presents an approach to the joint optimization of neural network structure and weights which can take advantage of backpropagation as a specialized decoder. The approach has been applied to a financial problem, whereby a factor model capturing the mutual relationships among several financial instruments is sought for. A sample application of such a model to statistical arbitrage is also presented.

Dermatology Disease Classification via Novel Evolutionary Artificial Neural Network

“…The neuro-genetic approach considered in this application has been presented in the literature and it has been validated on different benchmarks and real-world problems [2,1]. Our approach takes advantage of the Backpropagation (BP) as local search optimization algorithm.…”

“…A novel aspect of this approach is that, despite other previous approaches [2,1], the fitness of an individual is not given by the overall cost and the mean square error (mse) of each neural network, but based on the confusion matrix, in order to decrease the algorithm parameters and consequently reduce the fitness function complexity.…”

“…Here they are briefly presented, while a detailed description of such operators is reported in the literature [2,1].…”

“…• Selection The selection method implemented in this work follows the breeder genetic algorithm [1]. The selection strategy used by the algorithm is truncation, in which, starting from a population of n individuals, the worst ⌊n/2⌋ (with respect to f ) are eliminated.…”

“…The neurogenetic algorithm used in this setting has been previously presented in [2,1]. The novelty in this paper is that we carry out classification in presence of white noise in the dataset and implement a novel fitness function in the evolutionary process.…”