MNIST MLP¶
-
class
deepobs.tensorflow.testproblems.mnist_mlp.
mnist_mlp
(batch_size, weight_decay=None)[source]¶ DeepOBS test problem class for a multi-layer perceptron neural network on MNIST.
The network is build as follows:
- Four fully-connected layers with
1000
,500
,100
and10
units per layer. - The first three layers use ReLU activation, and the last one a softmax activation.
- The biases are initialized to
0.0
and the weight matrices with truncated normal (standard deviation of3e-2
) - The model uses a cross entropy loss.
- No regularization is used.
Parameters: - batch_size (int) -- Batch size to use.
- weight_decay (float) -- No weight decay (L2-regularization) is used in this
test problem. Defaults to
None
and any input here is ignored.
-
dataset
¶ The DeepOBS data set class for MNIST.
-
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. Will always be
0.0
since no regularizer is used.
-
accuracy
¶ A scalar tf.Tensor containing the mini-batch mean accuracy.
- Four fully-connected layers with