Looking at the training stage of error-backpropagation algorithm as an optimization problem, in this paper we check two ways of embedding simulated annealing to improve the usual gradient descent method for the achievement of good minima of the error function. The first way refers to a continuous state multilayer perceptron (MLP) and it stands for a random selection of the descent direction around the steepest one. The second way concerns binary states MLP where a backpropagation of the right answer from output to input is realized through a Boltzmann machine.
Simulated annealing approach in back propagation / S. AMATO, B. APOLLONI, G. CAPORALI, U. MADESANI, A.M. ZANABONI. - In: NEUROCOMPUTING. - ISSN 0925-2312. - 3:5-6(1991), pp. 207-220. [10.1016/0925-2312(91)90003-T]
Simulated annealing approach in back propagation
B. Apolloni;A.M. Zanaboni
1991
Abstract
Looking at the training stage of error-backpropagation algorithm as an optimization problem, in this paper we check two ways of embedding simulated annealing to improve the usual gradient descent method for the achievement of good minima of the error function. The first way refers to a continuous state multilayer perceptron (MLP) and it stands for a random selection of the descent direction around the steepest one. The second way concerns binary states MLP where a backpropagation of the right answer from output to input is realized through a Boltzmann machine.
Pubblicazioni consigliate
I metadati presenti in AIR sono rilasciati con licenza CCO mentre i file delle pubblicazioni sono rilasciati con licenza CC BY-SA, salvo diversa indicazione. In caso di violazione di copyright, contattare l'assistenza di AIR
