def affine_forward(x, w, b): out = None N = x.shape[0] x_row = x.reshape(N, -1) w = np.random.randn(784, 100) out = np.dot(x_row, w) + b.reshape(1, -1) cache = (x, w, b) return out, cache mnist识别shapes (40,6272) and (784,100) not aligned: 6272 (dim 1) != 784 (dim 0)如何解决

你需要将第一个矩阵的列数与第二个矩阵的行数相等，才能进行矩阵乘法。在这里，你需要将第一个矩阵的形状改为(40, 784)，可以通过x.reshape(N, -1)实现。修改后的代码如下：

def affine_forward(x, w, b):
    out = None
    N = x.shape[0]
    x_row = x.reshape(N, -1)
    out = np.dot(x_row, w) + b.reshape(1, -1)
    cache = (x, w, b)
    return out, cache

注意，这里将w的值重新赋值为np.random.randn(784, 100)是没有必要的，因为w是作为函数参数传入的，所以不需要重新赋值。