CIFAR-100 All-CNN-C

class deepobs.tensorflow.testproblems.cifar100_allcnnc.cifar100_allcnnc(batch_size, weight_decay=0.0005)

DeepOBS test problem class for the All Convolutional Neural Network C on Cifar-100.

Details about the architecture can be found in the original paper.

The paper does not comment on initialization; here we use Xavier for conv filters and constant 0.1 for biases.

A weight decay is used on the weights (but not the biases) which defaults to 5e-4.

The reference training parameters from the paper are batch size = 256, num_epochs = 350 using the Momentum optimizer with \(\mu = 0.9\) and an initial learning rate of \(\alpha = 0.05\) and decrease by a factor of 10 after 200, 250 and 300 epochs.

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 5e-4.

dataset

The DeepOBS data set class for Cifar-100.

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 All CNN C test problem on Cifar-100.