Fast learning for multi-layer perceptrons using statistical techniques

Author

Buhrke, Eric R. ; LoCicero, Joseph L.

Author_Institution

Dept. of Electr. & Comput. Eng., Illinois Inst. of Technol., Chicago, IL, USA

Volume

1

fYear

1992

fDate

23-26 Mar 1992

Firstpage

401

Abstract

The authors describe a new learning algorithm for the multi-layer perceptron. The learning problem is stated formally as an optimization problem that is solved through a series of systematic approximations. The solution uses the moments of the training data to design the network. This procedure has several advantages, most importantly the reduction in training time. The algorithm is verified and compared to backpropagation. In a speech recognition experiment the total training time was reduced by more than 75% when compared to backpropagation

Keywords

feedforward neural nets; learning (artificial intelligence); optimisation; speech recognition; statistical analysis; backpropagation; learning algorithm; learning problem; multi-layer perceptrons; optimization problem; speech recognition; statistical techniques; systematic approximations; training data moments; training time; Artificial neural networks; Backpropagation algorithms; Equations; Multi-layer neural network; Multilayer perceptrons; Network topology; Neural networks; Neurons; Speech recognition; Training data;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech, and Signal Processing, 1992. ICASSP-92., 1992 IEEE International Conference on

Conference_Location

San Francisco, CA

ISSN

1520-6149

Print_ISBN

0-7803-0532-9

Type

conf

DOI

10.1109/ICASSP.1992.225887

Filename

225887