多層感知機中：

hi 以 p 的概率被丟棄，以 1-p 的概率被拉伸，除以  1 - p

import mxnet as mx
import sys
import os
import time
import gluonbook as gb
from mxnet import autograd,init
from mxnet import nd,gluon
from mxnet.gluon import data as gdata,nn
from mxnet.gluon import loss as gloss


'''
# 模型引數
num_inputs, num_outputs, num_hiddens1, num_hiddens2 = 784,10,256,256

W1 = nd.random.normal(scale=0.01,shape=(num_inputs,num_hiddens1))
b1 = nd.zeros(num_hiddens1)

W2 = nd.random.normal(scale=0.01,shape=(num_hiddens1,num_hiddens2))
b2 = nd.zeros(num_hiddens2)

W3 = nd.random.normal(scale=0.01,shape=(num_hiddens2,num_outputs))
b3 = nd.zeros(num_outputs)

params = [W1,b1,W2,b2,W3,b3]

for param in params:
    param.attach_grad()

# 定義網路

 
'''
# 讀取資料
# fashionMNIST 28*28 轉為224*224
def load_data_fashion_mnist(batch_size, resize=None, root=os.path.join(
        '~', '.mxnet', 'datasets', 'fashion-mnist')):
    root = os.path.expanduser(root)  # 展開使用者路徑 '~'。
    transformer = []
    if resize:
        transformer += [gdata.vision.transforms.Resize(resize)]
    transformer 
 
+= [gdata.vision.transforms.ToTensor()]
    transformer = gdata.vision.transforms.Compose(transformer)
    mnist_train = gdata.vision.FashionMNIST(root=root, train=True)
    mnist_test = gdata.vision.FashionMNIST(root=root, train=False)
    num_workers = 0 if sys.platform.startswith('win32') else
 
 4
    train_iter = gdata.DataLoader(
        mnist_train.transform_first(transformer), batch_size, shuffle=True,
        num_workers=num_workers)
    test_iter = gdata.DataLoader(
        mnist_test.transform_first(transformer), batch_size, shuffle=False,
        num_workers=num_workers)
    return train_iter, test_iter


# 定義網路
drop_prob1,drop_prob2 = 0.2,0.5
# Gluon版
net = nn.Sequential()
net.add(nn.Dense(256,activation="relu"),
        nn.Dropout(drop_prob1),
        nn.Dense(256,activation="relu"),
        nn.Dropout(drop_prob2),
        nn.Dense(10)
        )
net.initialize(init.Normal(sigma=0.01))



# 訓練模型

def accuracy(y_hat, y):
    return (y_hat.argmax(axis=1) == y.astype('float32')).mean().asscalar()
def evaluate_accuracy(data_iter, net):
    acc = 0
    for X, y in data_iter:
        acc += accuracy(net(X), y)
    return acc / len(data_iter)


def train(net, train_iter, test_iter, loss, num_epochs, batch_size,
              params=None, lr=None, trainer=None):
    for epoch in range(num_epochs):
        train_l_sum = 0
        train_acc_sum = 0
        for X, y in train_iter:
            with autograd.record():
                y_hat = net(X)
                l = loss(y_hat, y)
            l.backward()
            if trainer is None:
                gb.sgd(params, lr, batch_size)
            else:
                trainer.step(batch_size)  # 下一節將用到。
            train_l_sum += l.mean().asscalar()
            train_acc_sum += accuracy(y_hat, y)
        test_acc = evaluate_accuracy(test_iter, net)
        print('epoch %d, loss %.4f, train acc %.3f, test acc %.3f'
              % (epoch + 1, train_l_sum / len(train_iter),
                 train_acc_sum / len(train_iter), test_acc))


num_epochs = 5
lr = 0.5
batch_size = 256
loss = gloss.SoftmaxCrossEntropyLoss()
train_iter, test_iter = load_data_fashion_mnist(batch_size)

trainer = gluon.Trainer(net.collect_params(),'sgd',{'learning_rate':lr})
train(net,train_iter,test_iter,loss,num_epochs,batch_size,None,None,trainer)