*NEURAL NETWORK

Define the configuration of a neural network.

This option is used to define the configuration of a neural network. It must be followed by a NEURAL NETWORK DATA option.

This page discusses:

See Also
*NEURAL NETWORK DATA
In Other Guides
About Neural Networks
Evaluating Abaqus Neural Networks

Products Abaqus/Standard

Type Model data

LevelModel

Required parameters

ID
Set this parameter equal to a unique integer value representing the neural network id.
NINPUT
Set this parameter equal to a positive, nonzero integer value representing the size (dimension) of the neural network input data array.
NOUTPUT
Set this parameter equal to a positive, nonzero integer value representing the size (dimension) of the neural network output data array. This value must be consistent with the size of the output (last) layer.
TYPE
Set this parameter equal to FFD, indicating a feed-forward, dense neural network definition.

Optional parameters

LABEL
Include this parameter to assign a label to the neural network.

Data lines to define the (feed-forward) neural network (TYPE=FFD)

First line
  1. Number of layers in the neural network.

Second line
  1. Size of the first layer in the neural network (integer value greater than zero)

  2. Set this entry to YES or NO to indicate whether or not the layer uses biases. If you set this entry to YES, you must provide the bias values for this layer under the subsequent NEURAL NETWORK DATA suboption. Otherwise, you must omit the bias values for this layer.

  3. Activation function type for this layer (CELU, ELU, FastTanh, Gauss, GELU, LeakyReLU, Linear, ReLU, Sigmoid, SiLU, Sin, SoftPlus, SoftSign, Swish, or Tanh).

  4. (Optional) Nondefault activation function parameter value if this layer uses the CELU, ELU, LeakyReLU, or Swish activation function.

Repeat this data line for every layer in the neural network.