| 摘要 |
PURPOSE: A multilayer perceptron structure is provided to increase the effectiveness of learning by modeling the form similar to brain cell of human, by differently applying weight values and forming neuron of each layer. CONSTITUTION: A signal of a specific level is inputted to neurons in first layer(31). The first neurons in second layer(32a) is formed with the first weight value in the neurons in first layer(31). The second neurons in second layer(32b) is formed with the second weight value in the neurons in first layer(31). A weight value adding unit(33) adds an output value of the first neurons in second layer(32a) and the second neurons in second layer(32b). The weight value adding unit(33) is formed and connected with the first neurons in second layer(32a) and the second neurons in second layer(32b).
|
