tfp.bijectors.masked_dense

A autoregressively masked dense layer. Analogous to tf.layers.dense.

tfp.bijectors.masked_dense(
    inputs,
    units,
    num_blocks=None,
    exclusive=False,
    kernel_initializer=None,
    reuse=None,
    name=None,
    *args,
    **kwargs
)

See [Germain et al. (2015)][1] for detailed explanation.

inputs Tensor input.
units Python int scalar representing the dimensionality of the output space.
num_blocks Python int scalar representing the number of blocks for the MADE masks.
exclusive Python bool scalar representing whether to zero the diagonal of the mask, used for the first layer of a MADE.
kernel_initializer Initializer function for the weight matrix. If None (default), weights are initialized using the tf.glorot_random_initializer.
reuse Python bool scalar representing whether to reuse the weights of a previous layer by the same name.
name Python str used to describe ops managed by this function.
*args tf.layers.dense arguments.
**kwargs tf.layers.dense keyword arguments.

Output tensor.

NotImplementedError if rightmost dimension of inputs is unknown prior to graph execution.

References

[1]: Mathieu Germain, Karol Gregor, Iain Murray, and Hugo Larochelle. MADE: Masked Autoencoder for Distribution Estimation. In International Conference on Machine Learning, 2015. https://arxiv.org/abs/1502.03509