PyTorch Deep Explainer MNIST 示例

一个简单的示例，展示如何使用 Deep Explainer 解释使用 PyTorch 训练的 MNIST CNN。

import numpy as np
import torch
from torch import nn, optim
from torch.nn import functional as F
from torchvision import datasets, transforms

import shap

batch_size = 128
num_epochs = 2
device = torch.device("cpu")


class Net(nn.Module):
    def __init__(self):
        super().__init__()

        self.conv_layers = nn.Sequential(
            nn.Conv2d(1, 10, kernel_size=5),
            nn.MaxPool2d(2),
            nn.ReLU(),
            nn.Conv2d(10, 20, kernel_size=5),
            nn.Dropout(),
            nn.MaxPool2d(2),
            nn.ReLU(),
        )
        self.fc_layers = nn.Sequential(
            nn.Linear(320, 50),
            nn.ReLU(),
            nn.Dropout(),
            nn.Linear(50, 10),
            nn.Softmax(dim=1),
        )

    def forward(self, x):
        x = self.conv_layers(x)
        x = x.view(-1, 320)
        x = self.fc_layers(x)
        return x


def train(model, device, train_loader, optimizer, epoch):
    model.train()
    for batch_idx, (data, target) in enumerate(train_loader):
        data, target = data.to(device), target.to(device)
        optimizer.zero_grad()
        output = model(data)
        loss = F.nll_loss(output.log(), target)
        loss.backward()
        optimizer.step()
        if batch_idx % 100 == 0:
            print(
                f"Train Epoch: {epoch} [{batch_idx * len(data)}/{len(train_loader.dataset)}"
                f" ({100.0 * batch_idx / len(train_loader):.0f}%)]"
                f"\tLoss: {loss.item():.6f}"
            )


def test(model, device, test_loader):
    model.eval()
    test_loss = 0
    correct = 0
    with torch.no_grad():
        for data, target in test_loader:
            data, target = data.to(device), target.to(device)
            output = model(data)
            test_loss += F.nll_loss(output.log(), target).item()  # sum up batch loss
            pred = output.max(1, keepdim=True)[
                1
            ]  # get the index of the max log-probability
            correct += pred.eq(target.view_as(pred)).sum().item()

    test_loss /= len(test_loader.dataset)
    print(
        f"\nTest set: Average loss: {test_loss:.4f},"
        f" Accuracy: {correct}/{len(test_loader.dataset)}"
        f" ({100.0 * correct / len(test_loader.dataset):.0f}%)\n"
    )


train_loader = torch.utils.data.DataLoader(
    datasets.MNIST(
        "mnist_data",
        train=True,
        download=True,
        transform=transforms.Compose([transforms.ToTensor()]),
    ),
    batch_size=batch_size,
    shuffle=True,
)

test_loader = torch.utils.data.DataLoader(
    datasets.MNIST(
        "mnist_data", train=False, transform=transforms.Compose([transforms.ToTensor()])
    ),
    batch_size=batch_size,
    shuffle=True,
)

model = Net().to(device)
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.5)

for epoch in range(1, num_epochs + 1):
    train(model, device, train_loader, optimizer, epoch)
    test(model, device, test_loader)

Train Epoch: 1 [0/60000 (0%)]   Loss: 2.311112
Train Epoch: 1 [12800/60000 (21%)]      Loss: 2.219999
Train Epoch: 1 [25600/60000 (43%)]      Loss: 1.546236
Train Epoch: 1 [38400/60000 (64%)]      Loss: 0.834946
Train Epoch: 1 [51200/60000 (85%)]      Loss: 0.731919

Test set: Average loss: 0.0046, Accuracy: 9006/10000 (90%)

Train Epoch: 2 [0/60000 (0%)]   Loss: 0.725186
Train Epoch: 2 [12800/60000 (21%)]      Loss: 0.456231
Train Epoch: 2 [25600/60000 (43%)]      Loss: 0.522802
Train Epoch: 2 [38400/60000 (64%)]      Loss: 0.553828
Train Epoch: 2 [51200/60000 (85%)]      Loss: 0.332361

Test set: Average loss: 0.0026, Accuracy: 9377/10000 (94%)

# since shuffle=True, this is a random sample of test data
batch = next(iter(test_loader))
images, _ = batch

background = images[:100]
test_images = images[100:103]

e = shap.DeepExplainer(model, background)
shap_values = e.shap_values(test_images)

2024-05-16 23:59:57.692986: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-05-16 23:59:58.335002: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT

shap_numpy = list(np.transpose(shap_values, (4, 0, 2, 3, 1)))
test_numpy = np.swapaxes(np.swapaxes(test_images.numpy(), 1, -1), 1, 2)

# plot the feature attributions
shap.image_plot(shap_numpy, -test_numpy)

上图显示了四个预测中每个类别的解释。请注意，解释是按照类别0-9从左到右沿着行排列的。