1_manual_gradient

手动梯度

w = 1.0


# 模型
class Model(object):
    def __init__(self):
        super(Model, self).__init__()

    # 前向传播函数
    def forward(self, x):
        return x * w

    # 损失函数
    def loss(self, x, y):
        y_pred = self.forward(x)
        return pow((y_pred - y), 2)

    # 计算梯度函数
    def gradient(self, x, y):
        return 2 * x * (x * w - y)

    def run(self):
        w = 0
        print('predict (before training)', 4, self.forward(4))
        for epoch in range(10):
            for x_val, y_val in zip(x_data, y_data):
                grad = self.gradient(x_val, y_val)
                w = w - 0.01 * grad
                print('\tgrad:', x_val, y_val, round(grad, 2))
                l = self.loss(x_val, y_val)
            print('process:', epoch, 'w=', round(w, 2), 'loss=', round(l, 2))

        print('predict (after training)', '4 hours', self.forward(4))


if __name__ == "__main__":
    print("Life is short, You need Python!")
    x_data = [1.0, 2.0, 3.0]
    y_data = [2.0, 4.0, 6.0]
    m = Model()
    m.run()