Graph

class torchhd.structures.Graph(dimensions: int, vsa: Literal['BSC', 'MAP', 'HRR', 'FHRR', 'BSBC', 'VTB', 'MCR', 'CGR'] = 'MAP', *, directed=False, device=None, dtype=None)

class torchhd.structures.Graph(input: VSATensor, *, directed=False)

Hypervector-based graph data structure.

Creates an empty sequence of dim dimensions or from an input tensor.

Parameters:

• dimensions (int) – number of dimensions of the graph.

• vsa – (VSAOptions, optional): specifies the hypervector type and operations used (Default: "MAP").

• directed (bool, optional) – specify if the graph is directed or not. Default: False.

• dtype (torch.dtype, optional) – the desired data type of returned tensor. Default: if None, uses a global default (see torch.set_default_tensor_type()).

• device (torch.device, optional) – the desired device of returned tensor. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.

• input (VSATensor) – tensor representing a graph hypervector.

Examples:

>>> G = structures.Graph(10000, directed=True)

add_edge(node1: VSATensor, node2: VSATensor) → None

Adds an edge to the graph.

If directed the direction goes from the first node to the second one.

Parameters:

• node1 (VSATensor) – Hypervector representing the first node of the edge.

• node2 (VSATensor) – Hypervector representing the second node of the edge.

Examples:

>>> letters = list(string.ascii_lowercase)
>>> letters_hv = torchhd.random(len(letters), 10000)
>>> G.add_edge(letters_hv[0], letters_hv[1])

clear() → None

Empties the graph.

Examples:

>>> G.clear()

contains(input: VSATensor) → Tensor

Returns the normalized dot similarity of the input vector against the graph.

Parameters:

input (Tensor) – Hypervector to compare against the multiset.

Examples:

>>> e = G.encode_edge(letters_hv[0], letters_hv[1])
>>> G.contains(e)
tensor(1.)

encode_edge(node1: VSATensor, node2: VSATensor) → VSATensor

Returns the encoding of an edge.

If directed the direction goes from the first node to the second one.

Parameters:

• node1 (VSATensor) – Hypervector representing the first node of the edge.

• node2 (VSATensor) – Hypervector representing the second node of the edge.

Examples:

>>> letters = list(string.ascii_lowercase)
>>> letters_hv = torchhd.random(len(letters), 10000)
>>> G.encode_edge(letters_hv[0], letters_hv[1])
tensor([-1.,  1., -1.,  ...,  1., -1., -1.])

classmethod from_edges(input: VSATensor, directed=False)

Creates a graph from a VSATensor

See: graph().

Parameters:

• input (VSATensor) – tensor containing pairs of node hypervectors that share an edge.

• directed (bool, optional) – specify if the graph is directed or not. Default: False.

Examples::

>>> edges = torch.tensor([[0, 0, 1, 2], [1, 2, 2, 3]])
>>> node_embedding = embeddings.Random(4, 10000)
>>> edges_hv = node_embedding(edges)
>>> graph = structures.Graph.from_edges(edges_hv)

node_neighbors(input: VSATensor, outgoing=True) → VSATensor

Returns the multiset of node neighbors of the input node.

Parameters:

• input (VSATensor) – Hypervector representing the node.

• outgoing (bool, optional) – if True, returns the neighboring nodes that input has an edge to. If False, returns the neighboring nodes that input has an edge from. This only has effect for directed graphs. Default: True.

Examples:

>>> G.node_neighbors(letters_hv[0])
tensor([ 1.,  1.,  1.,  ..., -1., -1.,  1.])