Java实现深度神经网络：Linear层

Java实现深度神经网络：Linear层

使用Java实现

Linear .java

/**
 * Linear 层
 * @author SXC 2020年9月11日 上午10:39:35
 */
public class Linear {// out_features个神经元
	double[][] weight;// 每行为一个神经元的weight
	double[] bias;
	int in_features, out_features;

	public Linear(int in_features, int out_features) {
		weight = new double[out_features][in_features];
		bias = new double[out_features];
		this.in_features = in_features;
		this.out_features = out_features;
	}

	public void setBias(double[] bias) {
		this.bias = bias;
		if (bias.length!=out_features) {
			System.out.println("偏置输入维数不符合要求，应输入维数："+out_features);
		}
	}

	public void setWeight(double[][] weight) {
		this.weight = weight;
		if (weight[0].length!=in_features) {
			System.out.println("权重输入列数不符合要求，应输入维数："+in_features);
		}
		if (weight.length!=out_features) {
			System.out.println("权重输入行数不符合要求，应输入维数："+out_features);
		}
	}

	public double[] getBias() {
		return bias;
	}

	public double[][] getWeight() {
		return weight;
	}

	public double[] out(double[] x) {
		double[] out = new double[out_features];
		for (int i = 0; i < out_features; i++) {
			for (int j = 0; j < in_features; j++) {
				out[i] += weight[i][j]*x[j];
			}
			out[i] += bias[i];
		}
		return out;
	}
}

测试代码

public class t {

	public static void main(String[] args) {
		Linear linear=new Linear(3, 2);
        double[][] weight= {{1,2,3},{3,4,3}};
        double[] bias= {9,5};
        linear.setBias(bias);
        linear.setWeight(weight);
        double[] x= {5,7,3};
        double[] out= linear.out(x);
        System.out.println(out[0]+"  " +out[1]);
	}

测试结果

37.0  57.0

使用Pytorch实现

import torch as t
linear=t.nn.Linear(3,2)
input=t.tensor([5,7,3],dtype=t.float32)
weight=t.tensor([[1,2,3],[3,4,3]],dtype=t.float32)
bias=t.tensor([9,5],dtype=t.float32)
linear.weight=t.nn.Parameter(weight)
linear.bias=t.nn.Parameter(bias)
print(linear(input))

运行结果

H:\ProgramData\Anaconda3\python.exe D:/PycharmProjects/untitled/t.py
tensor([37., 57.], grad_fn=<AddBackward0>)

Process finished with exit code 0

结果相同，达到目标