Title :
Computation of two-layer perceptron networks’ sensitivity to input perturbation
Author :
Yang, Jing ; Zeng, Xiao-qin ; Ng, Wing W Y ; Yeung, Daniel S.
Author_Institution :
Dept. of Comput. Sci. & Eng., Hohai Univ., Nanjing
Abstract :
The sensitivity of a neural networkpsilas output to its input perturbation is an important measure for evaluating the networkpsilas performance. In this paper we propose a novel method to quantify the sensitivity of a two-layer perceptron network (TLPN). The sensitivity is defined as the mathematical expectation of absolute output deviations due to input perturbations with respect to all possible inputs. In our method a bottom-up way is followed, in which the sensitivity of a neuron is first considered and then is that of the entire network. The main contribution of the method is that it requests a weak assumption on the input, that is its elements need only to be independent identically distributed, and thus is more practical to real applications. Some experiments have been conducted, and the results demonstrate high accuracy and efficiency of the method.
Keywords :
perceptrons; perturbation theory; absolute output deviations; input perturbation; mathematical expectation; network performance evaluation; neural network output sensitivity; two-layer perceptron networks; Computer networks; Computer science; Cybernetics; Laboratories; Machine learning; Mathematical model; Multilayer perceptrons; Neural networks; Neurons; Stochastic processes; Central Limit Theorem; Sensitivity; Two-Layer Perceptron Network;
Conference_Titel :
Machine Learning and Cybernetics, 2008 International Conference on
Conference_Location :
Kunming
Print_ISBN :
978-1-4244-2095-7
Electronic_ISBN :
978-1-4244-2096-4
DOI :
10.1109/ICMLC.2008.4620506