The training of neural networks is considered as a combinatorial optimization task and solved with the Reactive Tabu Search (RTS) method. RTS is applicable to non-differentiable functions, is robust with respect to the random initialization and effective in exploring the search space. A special-purpose VLSI architecture was developed to take advantage of the limited memory and processing requirements of RTS. A brief description of the chip and a test on a recurrent network are presented
Combinatiorial Optimization for Neural Nets: RTS Algorithm and Silicon
1994-01-01
Abstract
The training of neural networks is considered as a combinatorial optimization task and solved with the Reactive Tabu Search (RTS) method. RTS is applicable to non-differentiable functions, is robust with respect to the random initialization and effective in exploring the search space. A special-purpose VLSI architecture was developed to take advantage of the limited memory and processing requirements of RTS. A brief description of the chip and a test on a recurrent network are presentedFile in questo prodotto:
Non ci sono file associati a questo prodotto.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
