用pytorch1.0搭建簡單的神經網絡：進行多分類分析

用pytorch1.0搭建簡單的神經網絡：進行多分類分析

import torch
import torch.nn.functional as F # 包含激勵函數
import matplotlib.pyplot as plt

# 假數據
# make fake data
n_data = torch.ones(100, 2)
x0 = torch.normal(2*n_data, 1)      # class0 x data (tensor), shape=(100, 2)
y0 = torch.zeros(100)               # class0 y data (tensor), shape=(100, 1)
x1 = torch.normal(-2*n_data, 1)     # class1 x data (tensor), shape=(100, 2)
y1 = torch.ones(100)                # class1 y data (tensor), shape=(100, 1)
# 注意 x, y 數據的數據形式是一定要像下面一樣 (torch.cat 是合並數據)
x = torch.cat((x0, x1), 0).type(torch.FloatTensor)  # shape (200, 2) FloatTensor = 32-bit floating
y = torch.cat((y0, y1), ).type(torch.LongTensor)    # shape (200,) LongTensor = 64-bit integer

# The code below is deprecated in Pytorch 0.4. Now, autograd directly supports tensors
# x, y = Variable(x), Variable(y)
# 畫散點圖
plt.scatter(x.data.numpy()[:, 0], x.data.numpy()[:, 1], c=y.data.numpy(), s=100, lw=0, cmap='RdYlGn')
plt.show()

# 建立神經網絡
# 先定義所有的層屬性(__init__()), 然后再一層層搭建(forward(x))層於層的關系鏈接
class Net(torch.nn.Module):
    def __init__(self, n_feature, n_hidden, n_output):
        super(Net, self).__init__()  # 繼承 __init__ 功能
        # 定義每層用什么樣的形式
        self.hidden = torch.nn.Linear(n_feature, n_hidden)   # hidden layer
        self.out = torch.nn.Linear(n_hidden, n_output)   # output layer

    def forward(self, x):  # 這同時也是 Module 中的 forward 功能
        # 正向傳播輸入值, 神經網絡分析出輸出值
        x = F.relu(self.hidden(x))      # activation function for hidden layer
        x = self.out(x)
        return x

net = Net(n_feature=2, n_hidden=10, n_output=2)     # define the network
print(net)  # net architecture  == 顯示神經網絡結構
# Net(
#   (hidden): Linear(in_features=2, out_features=10, bias=True)
#   (out): Linear(in_features=10, out_features=2, bias=True)
# )
# 搭建完神經網絡后，對 神經網路參數（net.parameters()） 進行優化
# (1.選擇優化器 optimizer 是訓練的工具
optimizer = torch.optim.SGD(net.parameters(), lr=0.02)  # 傳入 net 的所有參數, 學習率
# (2.選擇優化的目標函數
loss_func = torch.nn.CrossEntropyLoss()  # the target label is NOT an one-hotted

plt.ion()   # something about plotting
# (3.開始訓練網絡
for t in range(100):
    out = net(x)                 # input x and predict based on x  # 喂給 net 訓練數據 x, 輸出預測值
    loss = loss_func(out, y)     # must be (1. nn output, 2. target), the target label is NOT one-hotted  # 計算兩者的誤差

    optimizer.zero_grad()   # clear gradients for next train      # 清空上一步的殘余更新參數值
    loss.backward()         # backpropagation, compute gradients  # 誤差反向傳播, 計算參數更新值
    optimizer.step()        # apply gradients                     # 將參數更新值施加到 net 的 parameters 上

    if t % 2 == 0:
        # plot and show learning process
        plt.cla()
        # 過了一道 softmax 的激勵函數后的最大概率才是預測值
        prediction = torch.max(out, 1)[1]
        pred_y = prediction.data.numpy()
        target_y = y.data.numpy()
        plt.scatter(x.data.numpy()[:, 0], x.data.numpy()[:, 1], c=pred_y, s=100, lw=0, cmap='RdYlGn')
        accuracy = float((pred_y == target_y).astype(int).sum()) / float(target_y.size) # 預測中有多少和真實值一樣
        plt.text(1.5, -4, 'Accuracy=%.2f' % accuracy, fontdict={'size': 20, 'color':  'red'})
        plt.pause(0.1)

plt.ioff()
plt.show()