Java实现深度神经网络：PReLU层

使用Java实现

/**
 * PReLU层
 * 
 * @author xiaocheng
 *
 */
public class PReLU {
	public PReLU() {

	}

	public double[] out(double[] x) {
		double[] out = new double[x.length];
		for (int i = 0; i < x.length; i++) {
			out[i] = PReLUF(x[i]);
		}
		return out;
	}

	private double PReLUF(double x) {
		if (x < 0)
			return x * 0.25f;
		return x;
	}
}

测试

public class Welcome {
	public static void main(String[] args) {
		double[] x = { -10, 7, 3 };
		PReLU sigmoid = new PReLU();
		double[] out1 = sigmoid.out(x);
		System.out.println(out1[0] + "  " + out1[1] + "  " + out1[2]);
	}
}

运行结果

-2.5  7.0  3.0

使用Pytorch实现

import torch as t
input=t.tensor([-10,7,3],dtype=t.float32)
prelu=t.nn.PReLU()
print(prelu(input))

运行结果

H:\ProgramData\Anaconda3\python.exe D:/PycharmProjects/untitled3/test.py
tensor([-2.5000,  7.0000,  3.0000], grad_fn=<PreluBackward>)

Process finished with exit code 0

结果相同，达到目标