vae.layers.dense¶
-
vae.layers.dense(inputs, units, activation=None, weight_normalization=False, init=False, kernel_initializer=<function variance_scaling_initializer.<locals>._initializer>, bias_initializer=<tensorflow.python.ops.init_ops.Zeros object>, scale_init=1.0, bias_init=0.0, name='Dense', reuse=False)¶ Dense layer with optional weight normalization
Parameters: - inputs :
tf.Tensor Inputs to the layer.
- units : int
Number of units (outputs).
- activation: optional (default=None)
Activation function to use for the layer.
- weight_normalization : bool, optional (default=False)
Whether to use weight normalization.
- init : bool, optional (default=False)
Whether to perform data-dependent initialization of parameters.
- kernel_initializer : optional
Initializer for weights. Defaults to Xavier initialization.
- bias_initializer : optional (default=tf.zeros_initializer())
Initializer for biases. Ignored when using data-dependent initialization with weight normalization.
- scale_init :
tf.Tensor, optional (default=1.0) Extra scaling to apply to data-dependent initialization for weight normalization.
- bias_init :
tf.Tensor, optional (default=0.0) Extra bias to apply to data-dependent initialization for weight normalization.
- name : optional (default=’Dense’)
Name of the layer.
- reuse : bool, optional (default=False)
Whether to reuse variables.
Returns: - outputs :
tf.Tensor Outputs from the layer.
- inputs :