Home
Scholarly Works
Paralysis free fast learning: smart neural nets
Conference

Paralysis free fast learning: smart neural nets

Abstract

Fast learning fully connected feed forward smart neural nets are introduced by avoiding the use of sigmoid non-linear function driven conventional neurons. This is achieved through the introduction of 'smart neuron'. By comparing the performance of the smart neural nets to the conventional neural nets, it is found that smart neural nets learn extremely faster. In addition, the sigmoid non-linear function driven conventional neurons are found to be not important in building feed forward neural nets.

Authors

Dahanayake BW; Upton ARM

Volume

3

Pagination

pp. 1527-1534

Publication Date

January 1, 1995

Conference proceedings

IEEE International Conference on Fuzzy Systems

Contact the Experts team