53 lines
2.1 KiB
Java
53 lines
2.1 KiB
Java
package org.dromara.easyai.rnnNerveEntity;
|
||
|
||
import org.dromara.easyai.matrixTools.Matrix;
|
||
import org.dromara.easyai.i.ActiveFunction;
|
||
import org.dromara.easyai.i.OutBack;
|
||
|
||
import java.util.Map;
|
||
|
||
/**
|
||
* @author lidapeng
|
||
* 隐层神经元
|
||
* @date 9:30 上午 2019/12/21
|
||
*/
|
||
public class HiddenNerve extends Nerve {
|
||
private final int depth;//所处深度
|
||
|
||
public HiddenNerve(int id, int depth, int upNub, int downNub, float studyPoint,
|
||
boolean init, ActiveFunction activeFunction, int rzType, float lParam, int rnnOutNumber) throws Exception {//隐层神经元
|
||
super(id, upNub, "HiddenNerve", downNub, studyPoint,
|
||
init, activeFunction, rzType, lParam, rnnOutNumber);
|
||
this.depth = depth;
|
||
}
|
||
|
||
@Override
|
||
public void input(long eventId, float parameter, boolean isKernelStudy, Map<Integer, Float> E
|
||
, OutBack outBack, boolean isEmbedding, Matrix rnnMatrix) throws Exception {//接收上一层的输入
|
||
boolean allReady = insertParameter(eventId, parameter, isEmbedding);
|
||
if (allReady) {//参数齐了，开始计算 sigma - threshold
|
||
if (isEmbedding && !isKernelStudy) {
|
||
outBack.getWordVector(getId(), getWOne(eventId));
|
||
destoryParameter(eventId);
|
||
} else {
|
||
float sigma = calculation(eventId, isEmbedding);
|
||
float out = activeFunction.function(sigma);//激活函数输出数值
|
||
if (rnnMatrix != null) {//rnn 1改输出值，2查看是否需要转向
|
||
out = out + rnnMatrix.getNumber(depth, getId() - 1);
|
||
}
|
||
if (isKernelStudy) {
|
||
outNub = out;
|
||
} else {
|
||
destoryParameter(eventId);
|
||
}
|
||
if (rnnMatrix != null && rnnMatrix.getX() == depth + 1) {//转向输出
|
||
sendRnnMessage(eventId, out, isKernelStudy, E, outBack, false, rnnMatrix);
|
||
} else {
|
||
sendMessage(eventId, out, isKernelStudy, E, outBack, false, rnnMatrix);
|
||
}
|
||
}
|
||
}
|
||
}
|
||
|
||
}
|