pyjuice.nodes

Nodes

`CircuitNodes`	The base class for a vector (block) of circuit nodes that share the same scope and structure.
`InputNodes`	A class representing vectors of input nodes.
`ProdNodes`	A class representing vectors of product nodes.
`SumNodes`	A class representing vectors of sum nodes.

`foreach`	Traverse all nodes of a PC and can a specific function.
`foldup_aggregate`	Traverse all nodes of a PC bottom-up and aggregate a per-node object.

Helpers that build structured (block-diagonal or block-sparse) edge patterns, passed as the edge_ids argument of summate().

`edge_constructors.block_diagonal_edge_constructor`	An edge constructor that connects sum and child node blocks in a one-to-one (block-diagonal) pattern: the i-th sum node block is connected only to the i-th child node block.
`edge_constructors.block_sparse_rnd_blk_edge_constructor`	An edge constructor that connects every sum node block to num_chs_per_block randomly chosen child node blocks, yielding a block-sparse (rather than fully-connected) sum layer.

`distributions.Bernoulli`	A class representing Bernoulli distributions.
`distributions.Categorical`	A class representing Categorical distributions.
`distributions.DiscreteLogistic`	A class representing Discrete Logistic distributions.
`distributions.Gaussian`	A class representing Gaussian distributions.
`distributions.MaskedCategorical`	A class representing Categorical distributions with masks.
`distributions.Literal`	A class representing Literal (indicator) distributions.
`distributions.Indicator`	A class representing Indicator distributions.
`distributions.SoftEvidenceCategorical`	A class representing a Categorical distribution that allows external soft evidence.
`distributions.SoftEvidenceIndicator`	A class representing an Indicator distribution that allows external soft evidence.
`distributions.External`	A class representing user-define distributions (PyJuice only processes the incoming log-probabilities).
`distributions.ExternProductCategorical`	A class representing a product distribution of two Categorical distributions: Pr_{A} * Pr_{B}.