nnet.java

robocode机器人

package apv.nrlibj;
import java.io.*;
import java.lang.*;
//import java.lang.Exception;
import java.util.*;
import java.util.zip.*;

import robocode.*;

/************************************************************************/
/*                                                                      */
/*                                                                      */
/*                 CLASS  NNet                                          */
/*                                                                      */
/*                                                                      */
/************************************************************************/

/**
* This Class (the most important) defines a Neural Network object.
* A NNet object contains a lot of computation nodes linked toghether in various
* ways but organized in layers.<BR>
* So, a NNet is an array of layers.<p>
* The computing phase proceeds from the layer 0 to last layer.<BR>
* For each layer:<BR>
* each node is computed appling its "trf" function.<BR>
* After that, before pass to next layer, again for each node is applied the
* "out" node function. This function moves the output value from the out-buffer
* to out variable and sincronize layer computation.
* (for the EBP phase, if used, the process is reversed using err
* variable).<p>
* A Node can be defined by user or can be one of predefined node type (NodeLin
* or NodeSigm). But a layer have same type nodes.<p>
* A layer can have a buffer. This buffer is another layer. A layer can have just
* one buffer but this buffer can have a buffer... in a chain way.
* When a layer have a buffer, its output values are copied into the input
* variable of destination layer nodes.<BR>
* These values will be computed when this buffer will be computed. So, if this
* buffer have a number &lt than the buffered layer these values will be considered
* in next forward cycle as a memory, otherwise in the same computational cycle.
* <PRE>
*         layer 0              layer 2       layer 3
*          |--|                 |--|          |--|
*          |  |---------------->|  |--------->|  |
*          |  |             |-->|  |          |  |
*          |  |        |--| |   |--|          |--|
*          |--|        |  |-|