【JAVA】批量梯度下降

/**
 * 批量梯度下降
 */
public class BatchGradient {
	public void batchGradientDescent() {
		double inputDataMatrix[][] = { { 1, 4 }, { 2, 5 }, { 5, 1 }, { 4, 2 } }; // X输入
		double expectResult[] = { 19, 26, 19, 20 }; // 期望输出值
		double w[] = { 2, 6 }; // 权重参数 因为这里只涉及到两个变量 ，即X为两列输入
		double learningRate = 0.01;
		double loss = 100; // 损失值
		for (int i = 0; i < 100 && loss > 0.0001; i++) {
			double err_sum = 0;
			for (int j = 0; j < 4; j++) {
				double h = 0;
				for (int k = 0; k < 2; k++) {
					h = h + inputDataMatrix[j][k] * w[k];
				}
				err_sum = expectResult[j] - h;
				for (int k = 0; k < 2; k++) {
					w[k] = w[k] + learningRate * err_sum
							* inputDataMatrix[j][k]; // 权值每次改变的幅度，这个公式是通过梯度下降得到的
				}
			}
			System.out
					.println("此时的w权值为：" + "w0:" + w[0] + "---" + "w1:" + w[1]);
			double loss_sum = 0;
			for (int j = 0; j < 4; j++) {
				double sum = 0;
				for (int k = 0; k < 2; k++) {
					sum = sum + inputDataMatrix[j][k] * w[k];
				}
				loss_sum += (expectResult[j] - sum) * (expectResult[j] - sum);
			}
			System.out.println("loss:" + loss_sum);
		}
	}

	public static void main(String[] args) {
		BatchGradient bg = new BatchGradient();
		bg.batchGradientDescent();
	}
}

refer to:http://blog.csdn.net/abcjennifer/article/details/7716281

http://www.xatarena.cn/javajswz/20130402/1313.html