Pytorch with the MNIST Dataset - MINST
rpi.analyticsdojo.com
70. PyTorch Deep Explainer MNIST example#
A simple example showing how to explain an MNIST CNN trained using PyTorch with Deep Explainer.
Adopted from: https://www.kaggle.com/ceshine/pytorch-deep-explainer-mnist-example
70.1. Install the modified SHAP package#
!pip install https://github.com/ceshine/shap/archive/master.zip
Collecting https://github.com/ceshine/shap/archive/master.zip
Downloading https://github.com/ceshine/shap/archive/master.zip
| 74.1MB 155.9MB/s
Requirement already satisfied (use --upgrade to upgrade): shap==0.25.0 from https://github.com/ceshine/shap/archive/master.zip in /usr/local/lib/python3.6/dist-packages
Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from shap==0.25.0) (1.14.6)
Requirement already satisfied: scipy in /usr/local/lib/python3.6/dist-packages (from shap==0.25.0) (1.1.0)
Requirement already satisfied: scikit-learn in /usr/local/lib/python3.6/dist-packages (from shap==0.25.0) (0.20.3)
Requirement already satisfied: matplotlib in /usr/local/lib/python3.6/dist-packages (from shap==0.25.0) (3.0.3)
Requirement already satisfied: pandas in /usr/local/lib/python3.6/dist-packages (from shap==0.25.0) (0.22.0)
Requirement already satisfied: tqdm in /usr/local/lib/python3.6/dist-packages (from shap==0.25.0) (4.28.1)
Requirement already satisfied: ipython in /usr/local/lib/python3.6/dist-packages (from shap==0.25.0) (5.5.0)
Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib->shap==0.25.0) (2.3.1)
Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib->shap==0.25.0) (2.5.3)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib->shap==0.25.0) (1.0.1)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.6/dist-packages (from matplotlib->shap==0.25.0) (0.10.0)
Requirement already satisfied: pytz>=2011k in /usr/local/lib/python3.6/dist-packages (from pandas->shap==0.25.0) (2018.9)
Requirement already satisfied: prompt-toolkit<2.0.0,>=1.0.4 in /usr/local/lib/python3.6/dist-packages (from ipython->shap==0.25.0) (1.0.15)
Requirement already satisfied: pickleshare in /usr/local/lib/python3.6/dist-packages (from ipython->shap==0.25.0) (0.7.5)
Requirement already satisfied: decorator in /usr/local/lib/python3.6/dist-packages (from ipython->shap==0.25.0) (4.4.0)
Requirement already satisfied: pexpect; sys_platform != "win32" in /usr/local/lib/python3.6/dist-packages (from ipython->shap==0.25.0) (4.6.0)
Requirement already satisfied: pygments in /usr/local/lib/python3.6/dist-packages (from ipython->shap==0.25.0) (2.1.3)
Requirement already satisfied: setuptools>=18.5 in /usr/local/lib/python3.6/dist-packages (from ipython->shap==0.25.0) (40.9.0)
Requirement already satisfied: simplegeneric>0.8 in /usr/local/lib/python3.6/dist-packages (from ipython->shap==0.25.0) (0.8.1)
Requirement already satisfied: traitlets>=4.2 in /usr/local/lib/python3.6/dist-packages (from ipython->shap==0.25.0) (4.3.2)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.6/dist-packages (from python-dateutil>=2.1->matplotlib->shap==0.25.0) (1.11.0)
Requirement already satisfied: wcwidth in /usr/local/lib/python3.6/dist-packages (from prompt-toolkit<2.0.0,>=1.0.4->ipython->shap==0.25.0) (0.1.7)
Requirement already satisfied: ptyprocess>=0.5 in /usr/local/lib/python3.6/dist-packages (from pexpect; sys_platform != "win32"->ipython->shap==0.25.0) (0.6.0)
Requirement already satisfied: ipython-genutils in /usr/local/lib/python3.6/dist-packages (from traitlets>=4.2->ipython->shap==0.25.0) (0.2.0)
Building wheels for collected packages: shap
Building wheel for shap (setup.py) ... ?25ldone
?25h Stored in directory: /tmp/pip-ephem-wheel-cache-xug5_6wp/wheels/8a/28/17/098d434a3f59f8529cb0ea4729568482332eef9127589ae8a8
Successfully built shap
70.2. Proceed#
import torch, torchvision
from torchvision import datasets, transforms
from torch import nn, optim
from torch.nn import functional as F
import numpy as np
import shap
70.3. Set Parameters for Neural Network#
Convolutional Neural network followed by fully connected.
batch_size = 128
num_epochs = 2
device = torch.device('cpu')
class Net(nn.Module):
def __init__(self):
super(Net, self).__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('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
epoch, batch_idx * len(data), len(train_loader.dataset),
100. * batch_idx / len(train_loader), loss.item()))
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('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
test_loss, correct, len(test_loader.dataset),
100. * correct / len(test_loader.dataset)))
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)
0%| | 16384/9912422 [00:00<01:09, 142596.08it/s]
Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to mnist_data/MNIST/raw/train-images-idx3-ubyte.gz
9920512it [00:01, 5952554.71it/s]
Extracting mnist_data/MNIST/raw/train-images-idx3-ubyte.gz
32768it [00:00, 432289.05it/s]
0it [00:00, ?it/s]
Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to mnist_data/MNIST/raw/train-labels-idx1-ubyte.gz
Extracting mnist_data/MNIST/raw/train-labels-idx1-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to mnist_data/MNIST/raw/t10k-images-idx3-ubyte.gz
1654784it [00:00, 5457348.83it/s]
8192it [00:00, 175134.12it/s]
Extracting mnist_data/MNIST/raw/t10k-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to mnist_data/MNIST/raw/t10k-labels-idx1-ubyte.gz
Extracting mnist_data/MNIST/raw/t10k-labels-idx1-ubyte.gz
Processing...
Done!
Train Epoch: 1 [0/60000 (0%)] Loss: 2.302780
Train Epoch: 1 [12800/60000 (21%)] Loss: 2.191153
Train Epoch: 1 [25600/60000 (43%)] Loss: 1.284060
Train Epoch: 1 [38400/60000 (64%)] Loss: 0.900758
Train Epoch: 1 [51200/60000 (85%)] Loss: 0.818337
Test set: Average loss: 0.0050, Accuracy: 8891/10000 (89%)
Train Epoch: 2 [0/60000 (0%)] Loss: 0.652153
Train Epoch: 2 [12800/60000 (21%)] Loss: 0.740618
Train Epoch: 2 [25600/60000 (43%)] Loss: 0.725341
Train Epoch: 2 [38400/60000 (64%)] Loss: 0.542940
Train Epoch: 2 [51200/60000 (85%)] Loss: 0.454126
Test set: Average loss: 0.0029, Accuracy: 9300/10000 (93%)
# 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)
shap_numpy = [np.swapaxes(np.swapaxes(s, 1, -1), 1, 2) for s in shap_values]
test_numpy = np.swapaxes(np.swapaxes(test_images.numpy(), 1, -1), 1, 2)
# plot the feature attributions
shap.image_plot(shap_numpy, -test_numpy)
The plot above shows the explanations for each class on four predictions. Note that the explanations are ordered for the classes 0-9 going left to right along the rows.