Add files via upload (microsoft#454)

Ch7 Level3 Pytorch

B-教学案例与实践/B6-神经网络基本原理简明教程/SourceCode/ch07-LinearMultipleClassification/Level3_Pytorch.py

@@ -0,0 +1,100 @@
+import numpy as np
+import matplotlib.pyplot as plt
+from pathlib import Path
+import math
+from torch.utils.data import TensorDataset, DataLoader
+from HelperClass.NeuralNet_1_2 import *
+from HelperClass.Visualizer_1_0 import *
+from Level2_ShowMultipleResult import ShowData
+import torch.nn as nn
+import torch.nn.functional as F
+import torch
+from torch.optim import Adam
+import warnings
+warnings.filterwarnings('ignore')
+
+file_name = "../../data/ch07.npz"
+
+def ShowResult(W, B, X, Y, xt, yt):
+    fig = plt.figure(figsize=(6,6))
+    DrawThreeCategoryPoints(X[:,0], X[:,1], Y[:], xlabel="x1", ylabel="x2", show=False)
+
+    b13 = (B[0,0] - B[0,2])/(W[1,2] - W[1,0])
+    w13 = (W[0,0] - W[0,2])/(W[1,2] - W[1,0])
+
+    b23 = (B[0,2] - B[0,1])/(W[1,1] - W[1,2])
+    w23 = (W[0,2] - W[0,1])/(W[1,1] - W[1,2])
+
+    b12 = (B[0,1] - B[0,0])/(W[1,0] - W[1,1])
+    w12 = (W[0,1] - W[0,0])/(W[1,0] - W[1,1])
+
+    x = np.linspace(0,1,2)
+    y = w13 * x + b13
+    p13, = plt.plot(x,y,c='r')
+
+    x = np.linspace(0,1,2)
+    y = w23 * x + b23
+    p23, = plt.plot(x,y,c='b')
+
+    x = np.linspace(0,1,2)
+    y = w12 * x + b12
+    p12, = plt.plot(x,y,c='g')
+
+    plt.legend([p13,p23,p12], ["13","23","12"])
+    plt.axis([-0.1,1.1,-0.1,1.1])
+
+    DrawThreeCategoryPoints(xt[:,0], xt[:,1], yt[:], xlabel="x1", ylabel="x2", show=True, isPredicate=True)
+
+class Model(nn.Module):
+    def __init__(self, input_size, class_num):
+        super(Model, self).__init__()
+        self.fc = nn.Linear(input_size, class_num)
+    def forward(self, x):
+        x = self.fc(x)
+        x = F.softmax(x)
+        return x
+
+if __name__ == '__main__':
+    max_epoch = 500
+    num_category = 3
+    reader = DataReader_1_3(file_name)
+    reader.ReadData()
+    # show raw data before normalization
+    reader.NormalizeX()
+
+    num_input = 2       # input size
+    # get numpy form data
+    XTrain, YTrain = reader.XTrain, reader.YTrain - 1
+    torch_dataset = TensorDataset(torch.FloatTensor(XTrain), torch.LongTensor(YTrain.reshape(-1,)))
+    reader.ToOneHot(num_category, base=1)       # transform to one-hot
+    ShowData(reader.XRaw, reader.YTrain)
+    train_loader = DataLoader(          # data loader class
+        dataset=torch_dataset,
+        batch_size=32,
+        shuffle=False,
+    )
+    loss_func = nn.CrossEntropyLoss()
+    model = Model(num_input,num_category)
+    optimizer = Adam(model.parameters(), lr=1e-2)
+
+
+    e_loss = []     # mean loss at every epoch
+    for epoch in range(max_epoch):
+        b_loss = []     # mean loss at every batch
+        for step, (batch_x, batch_y) in enumerate(train_loader):
+            optimizer.zero_grad()
+            pred = model(batch_x)
+            loss = loss_func(pred,batch_y)
+            loss.backward()
+            optimizer.step()
+            b_loss.append(loss.cpu().data.numpy())
+        print("Epoch: %d, Loss: %.5f" % (epoch, np.mean(b_loss)))
+
+    xt_raw = np.array([5, 1, 7, 6, 5, 6, 2, 7]).reshape(4, 2)
+    xt = reader.NormalizePredicateData(xt_raw)
+    xt = torch.FloatTensor(xt)
+    output = model(xt).cpu().data.numpy()
+
+    ShowResult(model.fc.weight.data.numpy().transpose(), model.fc.bias.data.numpy().reshape(1, 3),
+               reader.XTrain, reader.YTrain, xt, output)
+