我在日本小帅哥那学习了GCN

事件起源

最近在研究GNN，看了些许GNN的东西，心想着光看不练门外汉啊！这可不行，于是我开始自己动手实现一个GCN识别，一想到整一个模型demo，那必少不了MINIST数据集，反正就移花接木大法（MINIST可能会想，我这么忙，真的屑屑你！）一开始想着自己整一个，但是还没开始我就陷入了沉思，MINIST一个图片数据怎么去变成图边数据，于是万能的百度指导我进入了人均星星星的知乎，在那里，我发现有个日本小帅哥（后文我都讲日帅）已经做过了，那不说了，作为能cv绝不手写的浑水摸鱼星人，开始了偶尔心血来潮的扒代码历程。
原代码传送门

扒代码历程

1. 图片数据变成节点和边

怎么把图片数据变成节点和边？我想了很久，看了日帅的代码（为啥是看代码，因为日文我看不懂(o^^o)）我豁然开朗，其把每张图片每个像素点想成Node，其邻近关系考虑为边，具体思路如下（以3*3的数据举例）：
第一步：阈值过滤（为啥这样叫，因为我喜欢）

通过设定一个阈值k，源代码为102，我们这里设置为2（不知道为啥选这个，有知道的嘛？），将小于k的变成-1，反之为1000；

第二步：padding（为啥这样叫，大家都这么叫.~.）
源代码将padding_width定为2，其实我想了一下1是不是也可以；

第三步：得到Nodes和Edge的信息
将array中的非-1标记为节点k，k=0,1,2,3…

最后保存每个节点的坐标作为Node feature，以及边信息，例如3节点坐标为（1，1）边为[(3,1),(3,2),(3,4),(3,5),(3,6)]，至此图片数据变成图数据（日帅给我的启发很大的，回头我继续思考一下）。代码添加注释如下：

import gzip
import numpy as npdata = 0# 读取gzip图片数据，转换图片格式
with gzip.open('data/train-images-idx3-ubyte.gz', 'rb') as f:data = np.frombuffer(f.read(), np.uint8, offset=16)data = data.reshape([-1, 28, 28])# 把28*28的数据中<102变成1，大于变成1000，为啥取102我也不知道，嘻嘻嘻。
data = np.where(data < 102, -1, 1000)
for e,imgtmp in enumerate(data):# 数组padding,其实我在考虑做padding为1是不是也行img = np.pad(imgtmp, [(2, 2), (2, 2)], "constant", constant_values=(-1))# node标记 0，1，2，3，4.......cnt = 0for i in range(2, 30):for j in range(2, 30):if img[i][j] == 1000:img[i][j] = cntcnt += 1# 记录边和节点信息edges = []nodes = np.zeros((cnt, 2))for i in range(2, 30):for j in range(2, 30):if img[i][j] == -1:continuefilter = img[i - 2:i + 3, j - 2:j + 3].flatten()# Node的八个方位filter1 = filter[[6, 7, 8, 11, 13, 16, 17, 18]]# 记录节点的坐标nodes[filter[12]][0] = i - 2nodes[filter[12]][1] = j - 2# 记录边for tmp in filter1:if not tmp == -1:edges.append([filter[12], tmp])# 保存节点数据和边数据np.save("data/graphs/" + str(e), edges)np.save("data/node_features/" + str(e),nodes)

2. 模型训练

这部分的话，就不过细讲了（主要是我也过细讲不了），大致分为三个部分：
第一部分：加载数据
加载labels和处理好的Nodes以及edge数据，也就是说自建数据集，代码如下：

def load_mnist_graph(data_size=60000):# 获取数据主函数data_list = []labels = 0with gzip.open('data/train-labels-idx1-ubyte.gz', 'rb') as f:labels = np.frombuffer(f.read(), np.uint8, offset=8)for i in range(data_size):edge = torch.tensor(np.load('data/graphs/' + str(i) + '.npy').T, dtype=torch.long)x = torch.tensor(np.load('data/node_features/' + str(i) + '.npy') / 28, dtype=torch.float)# 构建数据集d = Data(x=x, edge_index=edge.contiguous(), t=int(labels[i]))data_list.append(d)if i % 1000 == 999:print("\rData loaded " + str(i + 1), end="  ")print("Complete!")return data_list

第二部分：定义网络
按照自己的喜好定义就行，毕竟我电脑cpu那点算力，还不支持我随心所欲的训练，M1早点出GPU版吧，孩子顶不住了(>﹏<)，这里就按照日帅的来吧！

# 定义网络结构
class Net(torch.nn.Module):def __init__(self):super(Net, self).__init__()self.conv1 = GCNConv(2, 16)self.conv2 = GCNConv(16, 32)self.conv3 = GCNConv(32, 48)self.conv4 = GCNConv(48, 64)self.conv5 = GCNConv(64, 96)self.conv6 = GCNConv(96, 128)self.linear1 = torch.nn.Linear(128,64)self.linear2 = torch.nn.Linear(64,10)def forward(self, data):x, edge_index = data.x, data.edge_indexx = self.conv1(x, edge_index)x = F.relu(x)x = self.conv2(x, edge_index)x = F.relu(x)x = self.conv3(x, edge_index)x = F.relu(x)x = self.conv4(x, edge_index)x = F.relu(x)x = self.conv5(x, edge_index)x = F.relu(x)x = self.conv6(x, edge_index)x = F.relu(x)x, _ = scatter_max(x, data.batch, dim=0)x = self.linear1(x)x = F.relu(x)x = self.linear2(x)return x

第三部分：训练主函数
训练部分的参数，可以按照自己电脑的算力以及结果定，这里不做过多修改，cv浑水摸鱼星人只是觉得日帅写的很棒(o^^o)！

def main():# 训练主程序data_size = 60000train_size = 50000batch_size = 100epoch_num = 150# 数据获取mnist_list = load_mnist_graph(data_size=data_size)device = torch.device('cpu')model = Net().to(device)trainset = mnist_list[:train_size]optimizer = torch.optim.Adam(model.parameters())trainloader = DataLoader(trainset, batch_size=batch_size, shuffle=True)testset = mnist_list[train_size:]testloader = DataLoader(testset, batch_size=batch_size)criterion = nn.CrossEntropyLoss()history = {"train_loss": [],"test_loss": [],"test_acc": []}print("Start Train")# 训练部分model.train()for epoch in range(epoch_num):train_loss = 0.0for i, batch in enumerate(trainloader):batch = batch.to("cpu")optimizer.zero_grad()outputs = model(batch)loss = criterion(outputs, batch.t)loss.backward()optimizer.step()train_loss += loss.cpu().item()if i % 10 == 9:progress_bar = '[' + ('=' * ((i + 1) // 10)) + (' ' * ((train_size // 100 - (i + 1)) // 10)) + ']'print('\repoch: {:d} loss: {:.3f}  {}'.format(epoch + 1, loss.cpu().item(), progress_bar), end="  ")print('\repoch: {:d} loss: {:.3f}'.format(epoch + 1, train_loss / (train_size / batch_size)), end="  ")history["train_loss"].append(train_loss / (train_size / batch_size))correct = 0total = 0batch_num = 0loss = 0with torch.no_grad():for data in testloader:data = data.to(device)outputs = model(data)loss += criterion(outputs, data.t)_, predicted = torch.max(outputs, 1)total += data.t.size(0)batch_num += 1correct += (predicted == data.t).sum().cpu().item()history["test_acc"].append(correct / total)history["test_loss"].append(loss.cpu().item() / batch_num)endstr = ' ' * max(1, (train_size // 1000 - 39)) + "\n"print('Test Accuracy: {:.2f} %%'.format(100 * float(correct / total)), end='  ')print(f'Test Loss: {loss.cpu().item() / batch_num:.3f}', end=endstr)print('Finished Training')# 最终结果correct = 0total = 0with torch.no_grad():for data in testloader:data = data.to(device)outputs = model(data)_, predicted = torch.max(outputs, 1)total += data.t.size(0)correct += (predicted == data.t).sum().cpu().item()print('Accuracy: {:.2f} %%'.format(100 * float(correct / total)))

完整代码
完整代码，我做了些许简化和修改，提醒一下这里的代码我改成了cpu，有条件的大帅哥可以自行改为cuda，如下：

import numpy as np
import gzip
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch_geometric.data import Data
from torch_geometric.loader import DataLoader
from torch_geometric.nn import GCNConv
from torch_scatter import  scatter_maxdef load_mnist_graph(data_size=60000):# 获取数据主函数data_list = []labels = 0with gzip.open('data/train-labels-idx1-ubyte.gz', 'rb') as f:labels = np.frombuffer(f.read(), np.uint8, offset=8)for i in range(data_size):edge = torch.tensor(np.load('data/graphs/' + str(i) + '.npy').T, dtype=torch.long)x = torch.tensor(np.load('data/node_features/' + str(i) + '.npy') / 28, dtype=torch.float)# 构建数据集d = Data(x=x, edge_index=edge.contiguous(), t=int(labels[i]))data_list.append(d)if i % 1000 == 999:print("\rData loaded " + str(i + 1), end="  ")print("Complete!")return data_list# 定义网络结构
class Net(torch.nn.Module):def __init__(self):super(Net, self).__init__()self.conv1 = GCNConv(2, 16)self.conv2 = GCNConv(16, 32)self.conv3 = GCNConv(32, 48)self.conv4 = GCNConv(48, 64)self.conv5 = GCNConv(64, 96)self.conv6 = GCNConv(96, 128)self.linear1 = torch.nn.Linear(128,64)self.linear2 = torch.nn.Linear(64,10)def forward(self, data):x, edge_index = data.x, data.edge_indexx = self.conv1(x, edge_index)x = F.relu(x)x = self.conv2(x, edge_index)x = F.relu(x)x = self.conv3(x, edge_index)x = F.relu(x)x = self.conv4(x, edge_index)x = F.relu(x)x = self.conv5(x, edge_index)x = F.relu(x)x = self.conv6(x, edge_index)x = F.relu(x)x, _ = scatter_max(x, data.batch, dim=0)x = self.linear1(x)x = F.relu(x)x = self.linear2(x)return xdef main():# 训练主程序data_size = 60000train_size = 50000batch_size = 100epoch_num = 150# 数据获取mnist_list = load_mnist_graph(data_size=data_size)device = torch.device('cpu')model = Net().to(device)trainset = mnist_list[:train_size]optimizer = torch.optim.Adam(model.parameters())trainloader = DataLoader(trainset, batch_size=batch_size, shuffle=True)testset = mnist_list[train_size:]testloader = DataLoader(testset, batch_size=batch_size)criterion = nn.CrossEntropyLoss()history = {"train_loss": [],"test_loss": [],"test_acc": []}print("Start Train")# 训练部分model.train()for epoch in range(epoch_num):train_loss = 0.0for i, batch in enumerate(trainloader):batch = batch.to("cpu")optimizer.zero_grad()outputs = model(batch)loss = criterion(outputs, batch.t)loss.backward()optimizer.step()train_loss += loss.cpu().item()if i % 10 == 9:progress_bar = '[' + ('=' * ((i + 1) // 10)) + (' ' * ((train_size // 100 - (i + 1)) // 10)) + ']'print('\repoch: {:d} loss: {:.3f}  {}'.format(epoch + 1, loss.cpu().item(), progress_bar), end="  ")print('\repoch: {:d} loss: {:.3f}'.format(epoch + 1, train_loss / (train_size / batch_size)), end="  ")history["train_loss"].append(train_loss / (train_size / batch_size))correct = 0total = 0batch_num = 0loss = 0with torch.no_grad():for data in testloader:data = data.to(device)outputs = model(data)loss += criterion(outputs, data.t)_, predicted = torch.max(outputs, 1)total += data.t.size(0)batch_num += 1correct += (predicted == data.t).sum().cpu().item()history["test_acc"].append(correct / total)history["test_loss"].append(loss.cpu().item() / batch_num)endstr = ' ' * max(1, (train_size // 1000 - 39)) + "\n"print('Test Accuracy: {:.2f} %%'.format(100 * float(correct / total)), end='  ')print(f'Test Loss: {loss.cpu().item() / batch_num:.3f}', end=endstr)print('Finished Training')# 最终结果correct = 0total = 0with torch.no_grad():for data in testloader:data = data.to(device)outputs = model(data)_, predicted = torch.max(outputs, 1)total += data.t.size(0)correct += (predicted == data.t).sum().cpu().item()print('Accuracy: {:.2f} %%'.format(100 * float(correct / total)))if __name__ == '__main__':main()

结语

今天是521，我和日帅的约会让我在对图数据上的收获收益匪浅，也达到了自己动手完成一个小demo的目标，其实说实话，日系帅哥的颜值我还是很吃的，还能写代码的就更爱了！又是一个cv浑水摸鱼的一天万岁！