梯度下降算法(python)

# 梯度下降算法
x_data = [1.0, 2.0, 3.0]
y_data = [1.5, 3.0, 4.5]

#参数
w = 1.0

# 学习率
lr = 0.01

def forward(x):
    '''预测值'''
    return x*w

def cost(xs, ys):
    '''计算损失和'''
    cost = 0
    for x, y in zip(xs, ys):
        y_pred = forward(x)
        cost += (y_pred-y) ** 2
    return cost

def gradient(xs, ys):
    '''计算梯度下降值'''
    grad = 0
    for x, y in zip(xs, ys):
        grad += 2*x*(forward(x)-y)
    return grad / len(xs)

for epoch in range(100000):
    '''lr 为学习率 即步长'''
    cost_val = cost(x_data, y_data)
    grad_val = gradient(x_data, y_data)
    w -= lr * grad_val
    print("Epoch:", epoch, ' w=', w, ' loss=', cost_val)
print("Predict (after traning) ", 100, forward(100))