CIFAR-10 3c3d

class deepobs.tensorflow.testproblems.cifar10_3c3d.cifar10_3c3d(batch_size, weight_decay=0.002)

DeepOBS test problem class for a three convolutional and three dense layered neural network on Cifar-10.

The network consists of

• thre conv layers with ReLUs, each followed by max-pooling
• two fully-connected layers with 512 and 256 units and ReLU activation
• 10-unit output layer with softmax
• cross-entropy loss
• L2 regularization on the weights (but not the biases) with a default factor of 0.002

The weight matrices are initialized using Xavier initialization and the biases are initialized to 0.0.

A working training setting is batch size = 128, num_epochs = 100 and SGD with learning rate of 0.01.

Parameters:

• batch_size (int) -- Batch size to use.
• weight_decay (float) -- Weight decay factor. Weight decay (L2-regularization) is used on the weights but not the biases. Defaults to 0.002.

dataset

The DeepOBS data set class for Cifar-10.

train_init_op

A tensorflow operation initializing the test problem for the training phase.

train_eval_init_op

A tensorflow operation initializing the test problem for evaluating on training data.

test_init_op

A tensorflow operation initializing the test problem for evaluating on test data.

losses

A tf.Tensor of shape (batch_size, ) containing the per-example loss values.

regularizer

A scalar tf.Tensor containing a regularization term.

accuracy

A scalar tf.Tensor containing the mini-batch mean accuracy.

set_up()

Set up the vanilla CNN test problem on Cifar-10.