| 摘要 |
A neural network (1) comprises an input port (5) connected to an output port (6) by one or more paths, each of which comprises an alternating series of weights (3) and neurons (2). The weights (3) amplify passing signals by a strengh factor. The network (1) can be trained by finding a set of strength factor values for the weights (3) such that the network (1) produces the correct output pattern from a given input pattern. During training, a strength factor perturbating and refresh means (10) applies perturbations to the strength factors of weights (3) in the net work, and updates the values of the strength factors depending on the difference between signals appearing at the output port (6), for a given pair of input and training patterns, when the weight is perturbed, and when it is not. |
