Title :
Feedforward neural networks with improved insensitivity abilities
Author_Institution :
Dipt. di Elettronica e Inf., Politecnico di Milano, Italy
Abstract :
The paper studies the insensitivity of regression-type feedforward neural networks, i.e., the ability possessed by the model of providing a graceful loss in performance when affected by perturbations. Such ability is somehow related to the application and, in general, cannot be simply improved by acting on the obtained model with off-line transformations. The attention is focused an perturbations affecting the network´s weights. We identify the worst case perturbation and quantify its effect on the network output. Modifications of the training function are suggested to improve the overall insensitivity of the model
Keywords :
feedforward neural nets; learning (artificial intelligence); perturbation techniques; insensitivity abilities; network output; network weights; off-line transformations; overall insensitivity; perturbations; regression-type feedforward neural networks; training function; worst case perturbation; Computer networks; Fault tolerance; Feedforward neural networks; Function approximation; Least squares approximation; Loss measurement; Neural networks; Quantization; Sensitivity analysis; Structural engineering;
Conference_Titel :
Circuits and Systems, 1999. ISCAS '99. Proceedings of the 1999 IEEE International Symposium on
Conference_Location :
Orlando, FL
Print_ISBN :
0-7803-5471-0
DOI :
10.1109/ISCAS.1999.777583
