day37简单的神经网络@浙大疏锦行

day37简单的神经网络@浙大疏锦行

使用 sklearn 的 load_digits 数据集 (8x8 像素的手写数字) 进行 MLP 训练。

importtorchimporttorch.nnasnnimporttorch.optimasoptimfromsklearn.datasetsimportload_digitsfromsklearn.model_selectionimporttrain_test_splitfromsklearn.preprocessingimportMinMaxScalerimportnumpyasnpimportmatplotlib.pyplotasplt# 1. 加载数据digits=load_digits()X=digits.data y=digits.targetprint(f"数据形状:{X.shape}")print(f"标签形状:{y.shape}")# 查看一张图片plt.imshow(digits.images[0],cmap='gray')plt.title(f"Label:{y[0]}")plt.show()

数据形状: (1797, 64) 标签形状: (1797,)

# 2. 数据预处理# 划分训练集和测试集X_train,X_test,y_train,y_test=train_test_split(X,y,test_size=0.2,random_state=42)# 归一化scaler=MinMaxScaler()X_train=scaler.fit_transform(X_train)X_test=scaler.transform(X_test)# 转换为 TensorX_train=torch.FloatTensor(X_train)y_train=torch.LongTensor(y_train)X_test=torch.FloatTensor(X_test)y_test=torch.LongTensor(y_test)print("训练集 Tensor 形状:",X_train.shape)print("测试集 Tensor 形状:",X_test.shape)

训练集 Tensor 形状: torch.Size([1437, 64])

测试集 Tensor 形状: torch.Size([360, 64])

# 3. 定义模型classMLP(nn.Module):def__init__(self):super(MLP,self).__init__()# 输入层 64 (8*8像素) -> 隐藏层 32 -> 输出层 10 (0-9数字)self.fc1=nn.Linear(64,32)self.relu=nn.ReLU()self.fc2=nn.Linear(32,10)defforward(self,x):out=self.fc1(x)out=self.relu(out)out=self.fc2(out)returnout model=MLP()print(model)

MLP(

(fc1): Linear(in_features=64, out_features=32, bias=True) (relu): ReLU()

(fc2): Linear(in_features=32, out_features=10, bias=True)

)

# 4. 定义损失函数和优化器criterion=nn.CrossEntropyLoss()optimizer=optim.SGD(model.parameters(),lr=0.1)# 学习率稍微调大一点，或者增加epoch

# 5. 训练模型num_epochs=2000losses=[]forepochinrange(num_epochs):# 前向传播outputs=model(X_train)loss=criterion(outputs,y_train)# 反向传播和优化optimizer.zero_grad()loss.backward()optimizer.step()losses.append(loss.item())if(epoch+1)%100==0:print(f'Epoch [{epoch+1}/{num_epochs}], Loss:{loss.item():.4f}')

# 6. 可视化损失plt.plot(range(num_epochs),losses)plt.xlabel('Epoch')plt.ylabel('Loss')plt.title('Training Loss')plt.show()

# 7. 模型评估withtorch.no_grad():# 训练集准确率outputs_train=model(X_train)_,predicted_train=torch.max(outputs_train,1)accuracy_train=(predicted_train==y_train).sum().item()/y_train.size(0)# 测试集准确率outputs_test=model(X_test)_,predicted_test=torch.max(outputs_test,1)accuracy_test=(predicted_test==y_test).sum().item()/y_test.size(0)print(f'训练集准确率:{accuracy_train:.4f}')print(f'测试集准确率:{accuracy_test:.4f}')

@浙大疏锦行