Solving the TSP by the AALHNN algorithm.

Area of Science:

• Computational Intelligence
• Operations Research
• Artificial Neural Networks

Background:

• Traditional Hopfield neural networks (HNN) struggle with local minima when solving the Traveling Salesman Problem (TSP).
• The penalty method in HNNs limits convergence to efficient TSP solutions.

Purpose of the Study:

• To propose an Accelerated Augmented Lagrangian Hopfield Neural Network (AALHNN) algorithm.
• To address the limitations of traditional HNNs in solving the TSP.
• To ensure efficient and optimal convergence for TSP solutions.

Main Methods:

• Developed an AALHNN algorithm incorporating Lagrangian multipliers and augmented Lagrange functions.
• Modified Hopfield and Tank TSP models by integrating Lagrange and augmented Lagrange multipliers with path length functions.
• Implemented Nesterov acceleration technique for updating Lagrange multipliers.
• Introduced a second-order factor to stabilize the neural network dynamic model.

Main Results:

• The AALHNN algorithm demonstrated robust convergence and successfully escaped local minimum traps.
• Theoretical proofs confirmed that the algorithm yields optimal solutions for the initial problem.
• Simulation experiments consistently achieved approximate optimal solutions for the TSP.
• The AALHNN method provided superior TSP solutions compared to traditional HNNs.

Conclusions:

• The AALHNN algorithm effectively solves the TSP, avoiding local minima and ensuring efficient solutions.
• This improved method offers a more robust and effective approach to TSP optimization.
• The findings highlight the potential of AALHNN for complex combinatorial optimization problems.