Quantum inspired PSO for the optimization of simultaneous recurrent neural networks as MIMO learning systems

Area of Science:

• Artificial Intelligence
• Machine Learning
• Control Systems

Background:

• Training simultaneous recurrent neural networks (SRNs) for multiple-input-multiple-output (MIMO) systems presents significant challenges.
• Existing methods struggle to efficiently learn all outputs of complex MIMO systems simultaneously.

Purpose of the Study:

• To introduce a novel training algorithm, particle swarm optimization with quantum infusion (PSO-QI), for SRNs in MIMO systems.
• To present a two-step learning approach to enhance the effectiveness and accuracy of SRN training for MIMO systems.

Main Methods:

• Developed the particle swarm optimization with quantum infusion (PSO-QI) algorithm, integrating swarm intelligence and quantum principles.
• Implemented a two-step learning strategy: first, optimizing all SRN weights for overall error reduction, then fine-tuning output-specific weights.
• Applied the PSO-QI algorithm and two-step learning to train SRNs for a benchmark MIMO system and a power system monitoring application.

Main Results:

• The PSO-QI algorithm demonstrated superior performance in training SRNs for MIMO systems.
• The two-step learning approach significantly improved the learning effectiveness and accuracy of the SRNs.
• Successful application in both a benchmark MIMO system and a real-world power system monitoring design.

Conclusions:

• SRNs can effectively learn complex MIMO systems when trained with the proposed PSO-QI algorithm and two-step learning approach.
• The PSO-QI algorithm offers a robust and efficient solution for training SRNs in challenging MIMO applications.
• The study validates the potential of this combined approach for advanced system modeling and control.