tensortools.ops

fully_connected

fully_connected(inputs, units, use_bias=True, activation=None, keep_prob=<tf.Tensor 'default_keep_prob:0' shape=() dtype=float32>, initializer=None)

A fully connected layer with dropout included.

Arguments:

• inputs: The inputs in the shape of [batch_size, previous_layer_units]
• units: The number of units for the layer.
• use_bias: Whether to use bias.
• activation: The TensorFlow activation function to use.
• keep_prob: The keep probability to use for the dropout layer.
• initializer: The kernal initializer.

Returns:

flatten

flatten(tensor, has_batch=False)

Flattens a tensor.

Arguments:

• tensor: Tensor of shape [batch_size, dim1, dim2 ... dimn] or [dim1, dim2 ... dimn]
• has_batch: Whether or not the tensor has a batch dimension. If so, it is preserved.

Returns:

Tensor of shape [batch_size, dim1dim2...dimn] or [dim1dim2 ... dimn]