# 收集数据，载入数据

from matplotlib import pyplot from regex import W import numpy def get_data():     x = numpy.linspace(-1, 1, 100)     y = 4 * x**4 - 5 * x**3 + 3 * x**2 - 2 * x - 1     return x, y # x, y = get_data() # pyplot.scatter(x, y) # pyplot.show() # 创建模型/假设函数形式 def f(x):     global W     return W[0] * x**4 + W[1] * x**3 + W[2] * x**2 + W[3] * x + W[4] # W = numpy.ones(5) # yp = f(x) # pyplot.scatter(x, y) # pyplot.plot(x, yp) # pyplot.show() # 赋初值 # W = numpy.ones(5) # 计算误差 def loss(yp, y):     return numpy.sum((yp - y) ** 2) # 计算梯度 def gradient_w(x, y, yp):     global W     g_w = numpy.zeros_like(W)     for i in range(5):         g_w[i] = 2 * numpy.sum((yp - y) * x ** (4 - i))     return g_w # 更新权重 # 循环迭代 def train(x, y, epochs, lr):     global W     w_list = [W]     l_list = []     for epoch in range(epochs):         yp = f(x)         l = loss(yp, y)         g_w = gradient_w(x, y, yp)         W = W - lr * g_w         w_list.append(W)         l_list.append(l)     return w_list, l_list x, y = get_data() W = numpy.ones(5) lr = 0.001 epochs = 500 w_list, l_list = train(x, y, epochs, lr) print(l_list[-1]) pyplot.plot(l_list) pyplot.show() for i, W in enumerate(w_list):     yp = f(x)     pyplot.cla()     pyplot.scatter(x, y)     pyplot.plot(x, yp, color="red")     pyplot.title(f"{i}")     pyplot.draw()     pyplot.pause(0.1) pyplot.show()