线性回归 — 从0开始

错误修正前 from mxnet import ndarray from mxnet import autograd import matplotlib.pyplot as plt true_w = [-2, 5] true_b = 1.5 num_batch = 1000 X = ndarray.random_normal(shape=[num_batch, 2]) y = ndarray.ones(shape=[num_batch, 1]) y = X[:, 0]*true_w[0]+X[:, 1]*true_w[1]+true_b

其中groundtrue的写法一般写成true_x。如true_w 数值型的变量一般用num_x。训练集写成num_examples ndarray.random_normal生成正态分布 ndarray.ondes生成全1矩阵，上面可以不用给y初始化。

修订后代码 from mxnet import ndarray from mxnet import autograd import matplotlib.pyplot as plt true_w = [-2, 5] true_b = 1.5 num_inputs = 2 num_examples = 1000 X = ndarray.random_normal(shape=[num_examples, num_inputs]) y = X[:, 0]*true_w[0]+X[:, 1]*true_w[1]+true_b

print(X[0:5],y[0:5])

[[ 1.16307855 0.48380461] [ 0.29956347 0.15302546] [-1.16881478 1.55807102] [-0.54594457 -2.35562968] [ 0.54144025 2.67850637]] <NDArray 5x2 @cpu(0)> [ 1.59286594 1.66600037 11.627985 -9.18625927 13.80965233] <NDArray 5 @cpu(0)> plt.scatter(X[:, 1].asnumpy(), y.asnumpy()) plt.show()