Enum: NeuralNetworkArchitecture

Architectural families of neural networks enumerated across Clause 3.4 and Clause 5.12.1.

URI: iso22989:NeuralNetworkArchitecture

Permissible Values

Value	Meaning	Description
feed_forward	None	Feed-forward neural network with unidirectional information flow (Clause 3
recurrent	None	Recurrent neural network with feedback connections (Clause 3
long_short_term_memory	None	LSTM recurrent architecture mitigating short memory in plain RNNs (Clause 3
gated_recurrent_unit	None	GRU recurrent architecture, a simplified gating variant of LSTM
convolutional	None	Convolutional neural network using local receptive fields (Clause 3
transformer	None	Attention-based architecture used for sequence modelling
autoencoder	None	Encoder-decoder architecture trained to reconstruct inputs for representation...
generative_adversarial	None	Adversarial pairing of generator and discriminator networks
deep	None	Neural network with many hidden layers (deep learning, Clause 3

Slots

Name	Description
neural_network_architecture	Neural-network architecture, when algorithm_family is neural_network (Clauses...

In Subsets

• Terminology

Identifier and Mapping Information

Schema Source

• from schema: https://w3id.org/lmodel/iso22989

LinkML Source

name: NeuralNetworkArchitecture
description: Architectural families of neural networks enumerated across Clause 3.4
  and Clause 5.12.1.
in_subset:
- terminology
from_schema: https://w3id.org/lmodel/iso22989
rank: 1000
permissible_values:
  feed_forward:
    text: feed_forward
    description: Feed-forward neural network with unidirectional information flow
      (Clause 3.4.6).
  recurrent:
    text: recurrent
    description: Recurrent neural network with feedback connections (Clause 3.4.10).
  long_short_term_memory:
    text: long_short_term_memory
    description: LSTM recurrent architecture mitigating short memory in plain RNNs
      (Clause 3.4.7).
  gated_recurrent_unit:
    text: gated_recurrent_unit
    description: GRU recurrent architecture, a simplified gating variant of LSTM.
  convolutional:
    text: convolutional
    description: Convolutional neural network using local receptive fields (Clause
      3.4.2).
  transformer:
    text: transformer
    description: Attention-based architecture used for sequence modelling.
  autoencoder:
    text: autoencoder
    description: Encoder-decoder architecture trained to reconstruct inputs for representation
      learning.
  generative_adversarial:
    text: generative_adversarial
    description: Adversarial pairing of generator and discriminator networks.
  deep:
    text: deep
    description: Neural network with many hidden layers (deep learning, Clause 3.4.4).