The NeuralNet library provides support for the use of trained or initialized neural networks in modeling.

In the context of system modeling, a neural network can be seen as an input/output block. At the core of this library, we provide a multiple-input and multiple-output (MIMO) block which calls the net as an external object. As such, it suffices to provide information on the location of the net and the dimensions of the input and ouput to obtain a realization of the net in System Modeler. For details on how to create and train neural networks in the Wolfram Language, explore the documentation on Neural Networks in the Wofram Language.

The current supported format is Open Neural Network Exchange (ONNX), through the use of the ONNX Runtime library, version 1.16.3, and its C API. ONNX is an open format for machine learning models supported by various tools and frameworks. ONNX models can be found in repositories, can be converted from other formats, and they can be generated with the use of the Wolfram Language. To create a MIMO block in System Modeler from a net with the use of the Wolfram Language, you can use CreateSystemModel.