本篇筆記的數據和完整代碼：https://github.com/ChenWentai/PyTorch

本期筆記使用PyTorch搭建一個多層神經網絡，解決具體的分類問題。

1 數據準備

數據采用糖尿病分類數據集diabetes.csv。這是一個典型的分類問題數據，包含768個樣本，每個樣本的數據包含8個特征，分別代表受試者的不同身體指標，標簽為0或1，代表是否患有糖尿病。數據集示意圖如下：

圖1. 糖尿病數據集

import torch
import torch.nn.functional as F
import torch.nn.init as init
import math
import numpy as np
import pandas as pd
from torch.autograd import Variable
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
#從本地讀取數據
xy = pd.read_csv('./diabetes.csv').values
x = Variable(torch.from_numpy(xy[:,0:-1]))
y = Variable(torch.from_numpy(xy[:,-1]))
#劃分訓練數據和測試數據
x_train, x_test,y_train,y_test= train_test_split(x.numpy(),y.numpy(), test_size=0.3,random_state=2018)   
ss = StandardScaler()  
#特征歸一化
x_train = Variable(torch.tensor(ss.fit_transform(x_train)))
x_test = Variable(torch.tensor(ss.fit_transform(x_test)))
y_train = Variable(torch.tensor(y_train))
y_test = Variable(torch.tensor(y_test))

2 搭建并訓練神經網絡

class Model(torch.nn.Module):
    def __init__(self):
        super(Model, self).__init__()
        self.l1 = torch.nn.Linear(8, 10)  
        self.l2 = torch.nn.Linear(10, 10)
        self.l3 = torch.nn.Linear(10, 10)
        self.l4 = torch.nn.Linear(10, 1)
    def forward(self, x):
        x = F.tanh(self.l1(x.float()))
        x = F.tanh(self.l2(x))
        x = F.tanh(self.l3(x))
        x = self.l4(x)
        return x
model = Model()
model.apply(weights_init)
criterion = torch.nn.BCEWithLogitsLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=1e-1, momentum=0.9)
Loss = []
Acc = []
EPOCHS = 5000
for epoch in range(EPOCHS):
    y_pred = model(x_train)
    loss = criterion(y_pred, y_train.float().view(-1,1))
    preds = torch.tensor(y_pred >= 0)
    corrects = torch.sum(preds.byte() == y_train.view(-1,1).byte())
    acc = corrects.item()/len(x_train)
    if epoch%100 == 0:
        print("corrects:",corrects)
        print("epoch = {0}, loss = {1}, acc = {2}".format(epoch, loss, acc))
        Loss.append(loss)
        Acc.append(acc)
    optimizer.zero_grad()
    loss.backward()
    optimizer.step()

經訓練后，訓練集上的acc可以達到1.0.
作出loss和acc的圖像

plt.plot(range(len(Loss)), Loss)
plt.ylabel('loss')
plt.xlabel('epochs')
plt.show()
plt.plot(range(len(Acc)), Acc)
plt.ylabel('acc')
plt.xlabel('epochs')
plt.show()

loss

acc

3 測試數據

在測試集上的表現如下

y_pred = model(x_test)
preds = torch.tensor(y_pred >= 0)
corrects = torch.sum(preds.byte() == y_test.view(-1,1).byte())
acc = corrects.item()/len(x_test)
print("corrects:",corrects.numpy().item())
print("acc = {}".format(acc))

corrects: 160
acc = 0.6926406926406926