A Novel Swarm Exploring Varying Parameter Recurrent Neural Network for Solving Non-Convex Nonlinear Programming.

Area of Science:

• Computational Mathematics
• Artificial Intelligence
• Optimization Algorithms

Background:

• Non-convex nonlinear programming presents significant computational challenges.
• Existing methods often struggle with efficiency and accuracy in finding optimal solutions.

Purpose of the Study:

• To propose a novel method, the swarm exploring varying parameter recurrent neural network (SE-VPRNN), for efficient and accurate non-convex nonlinear programming.
• To enhance global searching capabilities through wavelet mutation.

Main Methods:

• Utilizing a varying parameter recurrent neural network to accurately search for local optimal solutions.
• Integrating a particle swarm optimization (PSO) framework for information exchange and updating particle velocities and positions.
• Applying wavelet mutation to enhance particle diversity and global searching ability.

Main Results:

• The SE-VPRNN method effectively solves non-convex nonlinear programming problems.
• Computer simulations confirm the method's effectiveness.
• The proposed method shows advantages in accuracy and convergence time compared to three existing algorithms.

Conclusions:

• The SE-VPRNN method offers an effective solution for non-convex nonlinear programming.
• The integration of recurrent neural networks and PSO with wavelet mutation improves performance.
• This approach provides a promising alternative for complex optimization problems.