ontolearn.nero_architectures

Classes

MAB	Multi-head Attention Block.
SAB	Self-Attention Block.
ISAB	Induced Self-Attention Block.
PMA	Pooling by Multihead Attention.
SetTransformer	Set Transformer architecture.
DeepSet	DeepSet neural architecture for set-based learning.
SetTransformerNet	Set Transformer based architecture.

Module Contents

class ontolearn.nero_architectures.MAB(dim_Q, dim_K, dim_V, num_heads, ln=False)

Bases: torch.nn.Module

Multi-head Attention Block.

dim_V

num_heads

fc_q

fc_k

fc_v

fc_o

forward(Q, K)

class ontolearn.nero_architectures.SAB(dim_in, dim_out, num_heads, ln=False)

Bases: torch.nn.Module

Self-Attention Block.

mab

forward(X)

class ontolearn.nero_architectures.ISAB(dim_in, dim_out, num_heads, num_inds, ln=False)

Bases: torch.nn.Module

Induced Self-Attention Block.

I

mab0

mab1

forward(X)

class ontolearn.nero_architectures.PMA(dim, num_heads, num_seeds, ln=False)

Bases: torch.nn.Module

Pooling by Multihead Attention.

S

mab

forward(X)

class ontolearn.nero_architectures.SetTransformer(dim_input, num_outputs, dim_output, num_inds=32, dim_hidden=128, num_heads=4, ln=False)

Bases: torch.nn.Module

Set Transformer architecture.

enc

dec

forward(X)

class ontolearn.nero_architectures.DeepSet(num_instances: int, num_embedding_dim: int, num_outputs: int)

Bases: torch.nn.Module

DeepSet neural architecture for set-based learning.

name = 'DeepSet'

num_instances

num_embedding_dim

num_outputs

embeddings

fc0

fc1

forward(xpos, xneg)

positive_expression_embeddings(tensor_idx_individuals: torch.LongTensor)

negative_expression_embeddings(tensor_idx_individuals: torch.LongTensor)

class ontolearn.nero_architectures.SetTransformerNet(num_instances: int, num_embedding_dim: int, num_outputs: int)

Bases: torch.nn.Module

Set Transformer based architecture.

name = 'ST'

num_instances

num_embedding_dim

num_outputs

embeddings

set_transformer_negative

set_transformer_positive

forward(xpos, xneg)