AML Convolutional Neural Network for MNIST¶
Data and Libraries¶
In [3]:
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
import matplotlib.pyplot as plt
import numpy, random
# set the PseudoRandom Generator Seeds for better reproducibility
# see here for more: https://pytorch.org/docs/stable/notes/randomness.html
torch.manual_seed(99)
random.seed(99)
numpy.random.seed(99)
# this 'device' will be used for training our model
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print(device)
Load the MNIST dataset¶
Observe that we set shuffle=True, which means that data is randomized
In [4]:
input_size = 28*28 # images are 28x28 pixels
output_size = 10 # there are 10 classes
train_loader = torch.utils.data.DataLoader(
datasets.MNIST('../data', train=True, download=True,
transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])),
batch_size=64, shuffle=True)
test_loader = torch.utils.data.DataLoader(
datasets.MNIST('../data', train=False, transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])),
batch_size=1000, shuffle=True)
In [5]:
# show some training images
plt.figure(figsize=(16, 4))
# fetch a batch of train images; RANDOM
image_batch, label_batch = next(iter(train_loader))
for i in range(20):
image = image_batch[i]
label = label_batch[i].item()
plt.subplot(2, 10, i + 1)
#image, label = train_loader.dataset.__getitem__(i)
plt.imshow(image.squeeze().numpy())
plt.axis('off')
plt.title(label)
Helper functions for training and testing¶
In [6]:
# function to count number of parameters
def get_n_params(model):
np=0
for p in list(model.parameters()):
np += p.nelement()
return np
accuracy_list = []
# we pass a model object to this trainer, and it trains this model for one epoch
def train(epoch, model):
model.train()
for batch_idx, (data, target) in enumerate(train_loader):
# send to device
data, target = data.to(device), target.to(device)
optimizer.zero_grad()
output = model(data)
#loss = F.nll_loss(output, target)
loss = F.cross_entropy(output, 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):
model.eval()
test_loss = 0
correct = 0
for data, target in test_loader:
# send to device
data, target = data.to(device), target.to(device)
output = model(data)
#test_loss += F.nll_loss(output, target, reduction='sum').item() # sum up batch loss
test_loss += F.cross_entropy(output, target, reduction='sum').item() # sum up batch loss
pred = output.data.max(1, keepdim=True)[1] # get the index of the max log-probability
correct += pred.eq(target.data.view_as(pred)).cpu().sum().item()
test_loss /= len(test_loader.dataset)
accuracy = 100. * correct / len(test_loader.dataset)
accuracy_list.append(accuracy)
print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
test_loss, correct, len(test_loader.dataset),
accuracy))
The Convolutional Network¶
In [7]:
class CNN(nn.Module):
def __init__(self, input_size, output_size):
super(CNN, self).__init__()
self.conv1 = nn.Conv2d(in_channels=1, out_channels=12, kernel_size=3,padding=0)
self.conv2 = nn.Conv2d(in_channels=12, out_channels=24, kernel_size=6,padding=0)
self.conv3 = nn.Conv2d(in_channels=24, out_channels=32, kernel_size=6,padding=0)
self.fc1 = nn.Linear(8*4*4, 200)
self.fc2 = nn.Linear(200, 10)
def forward(self, x, verbose=False):
x = self.conv1(x)
x = F.relu(x)
x = self.conv2(x)
x = F.relu(x)
x = F.max_pool2d(x, kernel_size=2)
x = self.conv3(x)
x = F.relu(x)
x = F.max_pool2d(x, kernel_size=2)
x = x.view(-1, 8*4*4)
x = self.fc1(x)
x = F.relu(x)
x = self.fc2(x)
x = F.log_softmax(x, dim=1)
return x
Train the Network¶
In [8]:
print("Training on ", device)
model_cnn = CNN(input_size, output_size)
model_cnn.to(device)
optimizer = optim.SGD(model_cnn.parameters(), lr=0.01, momentum=0.5)
print('Number of parameters: {}'.format(get_n_params(model_cnn)))
for epoch in range(0, 10):
train(epoch, model_cnn)
test(model_cnn)
Show some predictions of the test network¶
In [9]:
def visualize_pred(img, pred_prob, real_label):
''' Function for viewing an image and it's predicted classes.
'''
#pred_prob = pred_prob.data.numpy().squeeze()
fig, (ax1, ax2) = plt.subplots(figsize=(6,9), ncols=2)
ax1.imshow(img.numpy().squeeze())
ax1.axis('off')
pred_label = numpy.argmax(pred_prob)
ax1.set_title([real_label, pred_label])
ax2.barh(numpy.arange(10), pred_prob)
ax2.set_aspect(0.1)
ax2.set_yticks(numpy.arange(10))
ax2.set_yticklabels(numpy.arange(10))
ax2.set_title('Prediction Probability')
ax2.set_xlim(0, 1.1)
plt.tight_layout()
In [10]:
model_cnn.to('cpu')
# fetch a batch of test images
image_batch, label_batch = next(iter(test_loader))
# Turn off gradients to speed up this part
with torch.no_grad():
log_pred_prob_batch = model_cnn(image_batch)
for i in range(10):
img = image_batch[i]
real_label = label_batch[i].item()
log_pred_prob = log_pred_prob_batch[i]
# Output of the network are log-probabilities, need to take exponential for probabilities
pred_prob = torch.exp(log_pred_prob).data.numpy().squeeze()
visualize_pred(img, pred_prob, real_label)
Network with Dropout¶
In [11]:
class CNNDropout(nn.Module):
def __init__(self, input_size, output_size):
super(CNNDropout, self).__init__()
self.conv1 = nn.Conv2d(in_channels=1, out_channels=12, kernel_size=3,padding=0)
self.conv2 = nn.Conv2d(in_channels=12, out_channels=24, kernel_size=6,padding=0)
self.conv3 = nn.Conv2d(in_channels=24, out_channels=32, kernel_size=6,padding=0)
self.fc1 = nn.Linear(8*4*4, 200)
self.do1 = nn.Dropout2d(p=0.8)
self.fc2 = nn.Linear(200, 10)
def forward(self, x, verbose=False):
x = self.conv1(x)
x = F.relu(x)
x = self.conv2(x)
x = F.relu(x)
x = F.max_pool2d(x, kernel_size=2)
x = self.conv3(x)
x = F.relu(x)
x = F.max_pool2d(x, kernel_size=2)
x = x.view(-1, 8*4*4)
x = self.fc1(x)
x = self.do1(x)
x = F.relu(x)
x = self.fc2(x)
x = F.log_softmax(x, dim=1)
return x
In [12]:
print("Training on ", device)
model_2 = CNNDropout(input_size, output_size)
model_2.to(device)
optimizer = optim.SGD(model_2.parameters(), lr=0.01, momentum=0.5)
print('Number of parameters: {}'.format(get_n_params(model_2)))
for epoch in range(0, 10):
model_2.train() # model in training mode. Turns on dropout, batch-norm etc during training
train(epoch, model_2)
model_2.eval() # model in evaluation mode. Turn off dropout, batch-norm etc during validation/testing
test(model_2)
In [13]:
model_2.to('cpu')
# fetch a batch of test images
image_batch, label_batch = next(iter(test_loader))
# Turn off gradients to speed up this part
with torch.no_grad():
log_pred_prob_batch = model_2(image_batch)
for i in range(10):
img = image_batch[i]
real_label = label_batch[i].item()
log_pred_prob = log_pred_prob_batch[i]
# Output of the network are log-probabilities, need to take exponential for probabilities
pred_prob = torch.exp(log_pred_prob).data.numpy().squeeze()
visualize_pred(img, pred_prob, real_label)
Does the CNN use "Visual Information" ?¶
In [14]:
fixed_perm = torch.randperm(784) # Fix a permutation of the image pixels; We apply the same permutation to all images
# show some training images
plt.figure(figsize=(8, 8))
# fetch a batch of train images; RANDOM
image_batch, label_batch = next(iter(train_loader))
for i in range(6):
image = image_batch[i]
image_perm = image.view(-1, 28*28).clone()
image_perm = image_perm[:, fixed_perm]
image_perm = image_perm.view(-1, 1, 28, 28)
label = label_batch[i].item()
plt.subplot(3,4 , 2*i + 1)
#image, label = train_loader.dataset.__getitem__(i)
plt.imshow(image.squeeze().numpy())
plt.axis('off')
plt.title(label)
plt.subplot(3, 4, 2*i+2)
plt.imshow(image_perm.squeeze().numpy())
plt.axis('off')
plt.title(label)
In [15]:
accuracy_list = []
def scramble_train(epoch, model, perm=torch.arange(0, 784).long()):
model.train()
for batch_idx, (data, target) in enumerate(train_loader):
# send to device
data, target = data.to(device), target.to(device)
# permute pixels
data = data.view(-1, 28*28)
data = data[:, perm]
data = data.view(-1, 1, 28, 28)
optimizer.zero_grad()
output = model(data)
loss = F.nll_loss(output, 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 scramble_test(model, perm=torch.arange(0, 784).long()):
model.eval()
test_loss = 0
correct = 0
for data, target in test_loader:
# send to device
data, target = data.to(device), target.to(device)
# permute pixels
data = data.view(-1, 28*28)
data = data[:, perm]
data = data.view(-1, 1, 28, 28)
output = model(data)
test_loss += F.nll_loss(output, target, reduction='sum').item() # sum up batch loss
pred = output.data.max(1, keepdim=True)[1] # get the index of the max log-probability
correct += pred.eq(target.data.view_as(pred)).cpu().sum().item()
test_loss /= len(test_loader.dataset)
accuracy = 100. * correct / len(test_loader.dataset)
accuracy_list.append(accuracy)
print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
test_loss, correct, len(test_loader.dataset),
accuracy))
In [16]:
print("Training on ", device)
model_cnn_3 = CNN(input_size, output_size)
model_cnn_3.to(device)
optimizer = optim.SGD(model_cnn_3.parameters(), lr=0.01, momentum=0.5)
print('Number of parameters: {}'.format(get_n_params(model_cnn_3)))
for epoch in range(0, 10):
scramble_train(epoch, model_cnn_3, fixed_perm)
scramble_test(model_cnn_3, fixed_perm)
Performance decreased from 99% to 96%¶
In [17]:
model_cnn_3.to('cpu')
# fetch a batch of test images
image_batch, label_batch = next(iter(test_loader))
image_batch_scramble = image_batch.view(-1, 28*28)
image_batch_scramble = image_batch_scramble[:, fixed_perm]
image_batch_scramble = image_batch_scramble.view(-1, 1, 28, 28)
# Turn off gradients to speed up this part
with torch.no_grad():
log_pred_prob_batch = model_cnn_3(image_batch_scramble)
for i in range(10):
img = image_batch[i]
img_perm = image_batch_scramble[i]
real_label = label_batch[i].item()
log_pred_prob = log_pred_prob_batch[i]
# Output of the network are log-probabilities, need to take exponential for probabilities
pred_prob = torch.exp(log_pred_prob).data.numpy().squeeze()
visualize_pred(img_perm, pred_prob, real_label)
