#使用MXNet提供的Gluon接口方便的实现线性回归的训练

#1-生成数据集
from mxnet import autograd,nd
num_inputs = 2
num_examples = 1000
true_w = [2,-3.4]
true_b = 4.2
features = nd.random.normal(scale=1,shape=(num_examples,num_inputs))
labels = true_w[0]*features[:,0]+true_w[1]*features[:,1]+true_b
labels += nd.random.normal(scale=0.01,shape=labels.shape)

#2-读取数据集
#gluon提供了data包来读取数据
from mxnet.gluon import data as gdata
batch_size = 10
#将训练的特征和标签组合
dataset = gdata.ArrayDataset(features,labels)
#随机读取小批量
data_iter = gdata.DataLoader(dataset,batch_size,shuffle=True)

#3-定义模型
from mxnet.gluon import nn 
net = nn.Sequential() #串联各个层的容器
net.add(nn.Dense(1)) #单层神经网络，输出层为全连接层，输出个数为1

#4-初始化模型参数,init模块
from mxnet import init
net.initialize(init.Normal(sigma=0.01)) 
#权重参数为正态分布，偏差参数初始化为0

#5-定义损失函数
from mxnet.gluon import loss as gloss
loss = gloss.L2Loss()  #平方损失，L2范数损失

#6-定义优化算法
from mxnet import gluon
trainer = gluon.Trainer(net.collect_params(),'sgd',{'learning_rate':0.03})
#collect_params()获取通过add添加的每一层包含的参数，sgd-》随机梯度下降优化算法

#7-训练模型
num_epochs = 3
for epoch in range(1,num_epochs+1):
    for X,y in data_iter:
        with autograd.record():
            l = loss(net(X),y)
        l.backward()
        trainer.step(batch_size)
    l=loss(net(features),labels)
    print('epoch %d,loss %f'%(epoch,l.mean().asnumpy()))

epoch 1,loss 0.039983
epoch 2,loss 0.000155
epoch 3,loss 0.000047

dense = net[0]
true_w,dense.weight.data()
true_b,dense.bias.data()

(4.2,
 
 [4.199878]
 <NDArray 1 @cpu(0)>)