DOMAIN KNOWLEDGE ENCAPSULATION

Reusable and fine-tunable software or hardware intellectual property (IP) library components featuring single-stage or multi-stage neural network architectures that incorporate control and data path elements, including but not limited to finite state machines (FSMs) and arithmetic logic units (ALUs), wherein both the architectural structure (e.g., configuration of FSMs, ALUs, or network layers) is predefined and the network parameters (such as weights and biases) are pretrained to encapsulate domain-specific knowledge for use in control, simulation, modeling, interfacing, inference, or state or parameter estimation.

Design and development of artificial intelligence systems providing reusable and fine-tunable library components with design support, based on single-stage or multi-stage neural network architectures that incorporate control and data path elements, including but not limited to finite state machines (FSMs) and arithmetic logic units (ALUs), wherein both the architectural structure (e.g., configuration of FSMs, ALUs, or network layers) is predefined and the network parameters (such as weights and biases) are pretrained to encapsulate domain-specific knowledge for purposes of control, simulation, modeling, interfacing, inference, or state or parameter estimation.