#!/usr/bin/env python
-# @XREMOTE_HOST: elk.fleuret.org
-# @XREMOTE_EXEC: ~/conda/bin/python
-# @XREMOTE_PRE: ln -s ~/data/pytorch ./data
-# @XREMOTE_PRE: killall -q -9 python || echo "Nothing killed"
-
import math, sys, torch, torchvision
from torch import nn
######################################################################
-# Returns a pair of tensors (x, c), where x is a Nx2x28x28 containing
-# pairs of images of same classes (one per channel), and p is a 1d
-# long tensor with the count of pairs per class used
+train_set = torchvision.datasets.MNIST('./data/mnist/', train = True, download = True)
+train_input = train_set.train_data.view(-1, 1, 28, 28).float()
+train_target = train_set.train_labels
+
+test_set = torchvision.datasets.MNIST('./data/mnist/', train = False, download = True)
+test_input = test_set.test_data.view(-1, 1, 28, 28).float()
+test_target = test_set.test_labels
+
+mu, std = train_input.mean(), train_input.std()
+train_input.sub_(mu).div_(std)
+test_input.sub_(mu).div_(std)
+
+used_MNIST_classes = torch.tensor([ 0, 1, 3, 5, 6, 7, 8, 9])
+# used_MNIST_classes = torch.tensor([ 0, 9, 7 ])
+# used_MNIST_classes = torch.tensor([ 3, 4, 7, 0 ])
+
+######################################################################
+
+# Returns a triplet of tensors (a, b, c), where a and b contain each
+# half of the samples, with a[i] and b[i] of same class for any i, and
+# c is a 1d long tensor with the count of pairs per class used.
+
+def create_MNIST_pair_set(train = False):
+ ua, ub = [], []
-def create_pair_set(used_classes, input, target):
- u = []
+ if train:
+ input, target = train_input, train_target
+ else:
+ input, target = test_input, test_target
- for i in used_classes:
+ for i in used_MNIST_classes:
used_indices = torch.arange(input.size(0), device = target.device)\
.masked_select(target == i.item())
x = input[used_indices]
x = x[torch.randperm(x.size(0))]
- # Careful with odd numbers of samples in a class
- x = x[0:2 * (x.size(0) // 2)].reshape(-1, 2, 28, 28)
- u.append(x)
+ hs = x.size(0)//2
+ ua.append(x.narrow(0, 0, hs))
+ ub.append(x.narrow(0, hs, hs))
- x = torch.cat(u, 0).contiguous()
- c = torch.tensor([x.size(0) for x in u])
+ a = torch.cat(ua, 0)
+ b = torch.cat(ub, 0)
+ perm = torch.randperm(a.size(0))
+ a = a[perm].contiguous()
+ b = b[perm].contiguous()
+ c = torch.tensor([x.size(0) for x in ua])
- return x, c
+ return a, b, c
######################################################################
self.fc1 = nn.Linear(256, 200)
self.fc2 = nn.Linear(200, 1)
- def forward(self, x):
+ def forward(self, a, b):
+ # Make the two images a single two-channel image
+ x = torch.cat((a, b), 1)
x = F.relu(F.max_pool2d(self.conv1(x), kernel_size = 3))
x = F.relu(F.max_pool2d(self.conv2(x), kernel_size = 2))
x = x.view(x.size(0), -1)
######################################################################
-train_set = torchvision.datasets.MNIST('./data/mnist/', train = True, download = True)
-train_input = train_set.train_data.view(-1, 1, 28, 28).float()
-train_target = train_set.train_labels
-
-mu, std = train_input.mean(), train_input.std()
-train_input.sub_(mu).div_(std)
-
-######################################################################
-
-# The information bound is the log of the number of classes in there
-
-# used_classes = torch.tensor([ 0, 1, 3, 5, 6, 7, 8, 9])
-used_classes = torch.tensor([ 3, 4, 7, 0 ])
-
nb_epochs, batch_size = 50, 100
model = Net()
+
+print('nb_parameters %d' % sum(x.numel() for x in model.parameters()))
+
optimizer = torch.optim.Adam(model.parameters(), lr = 1e-3)
if torch.cuda.is_available():
model.cuda()
train_input, train_target = train_input.cuda(), train_target.cuda()
+ test_input, test_target = test_input.cuda(), test_target.cuda()
for e in range(nb_epochs):
- input, count = create_pair_set(used_classes, train_input, train_target)
+ input_a, input_b, count = create_MNIST_pair_set(train = True)
+
+ # The information bound is the entropy of the class distribution
class_proba = count.float()
class_proba /= class_proba.sum()
class_entropy = - (class_proba.log() * class_proba).sum().item()
- input = input[torch.randperm(input.size(0))]
- indep_input = input.clone()
- indep_input[:, 1] = input[torch.randperm(input.size(0)), 1]
+ input_br = input_b[torch.randperm(input_b.size(0))]
- mi = 0.0
+ acc_mi = 0.0
- for batch, indep_batch in zip(input.split(batch_size), indep_input.split(batch_size)):
- loss = - (model(batch).mean() - model(indep_batch).exp().mean().log())
- mi -= loss.item()
+ for batch_a, batch_b, batch_br in zip(input_a.split(batch_size),
+ input_b.split(batch_size),
+ input_br.split(batch_size)):
+ mi = model(batch_a, batch_b).mean() - model(batch_a, batch_br).exp().mean().log()
+ loss = - mi
+ acc_mi += mi.item()
optimizer.zero_grad()
loss.backward()
optimizer.step()
- mi /= (input.size(0) // batch_size)
+ acc_mi /= (input_a.size(0) // batch_size)
- print('%d %.04f %.04f'%(e, mi / math.log(2), class_entropy / math.log(2)))
+ print('%d %.04f %.04f' % (e, acc_mi / math.log(2), class_entropy / math.log(2)))
sys.stdout.flush()
######################################################################
+
+input_a, input_b, count = create_MNIST_pair_set(train = False)
+
+for e in range(nb_epochs):
+ class_proba = count.float()
+ class_proba /= class_proba.sum()
+ class_entropy = - (class_proba.log() * class_proba).sum().item()
+
+ input_br = input_b[torch.randperm(input_b.size(0))]
+
+ acc_mi = 0.0
+
+ for batch_a, batch_b, batch_br in zip(input_a.split(batch_size),
+ input_b.split(batch_size),
+ input_br.split(batch_size)):
+ mi = model(batch_a, batch_b).mean() - model(batch_a, batch_br).exp().mean().log()
+ acc_mi += mi.item()
+
+ acc_mi /= (input_a.size(0) // batch_size)
+
+print('test %.04f %.04f'%(acc_mi / math.log(2), class_entropy / math.log(2)))
+
+######################################################################
projects / pytorch.git / blobdiff