Hash Bit Selection Based on Collaborative Neurodynamic Optimization.

Area of Science:

• Computer Science
• Optimization Algorithms
• Machine Learning

Background:

• Hash bit selection is crucial for optimizing data structures and algorithms.
• Existing methods often struggle with the complexity of selecting optimal hash bits, which can be formulated as a quadratic programming problem.

Purpose of the Study:

• To develop and validate a new optimization approach for hash bit selection.
• To address the limitations of current methods in finding optimal hash bit subsets.

Main Methods:

• Reformulated the hash bit selection problem as a global optimization problem.
• Applied a Collaborative Neurodynamic Optimization (CNO) approach, initializing models with Particle Swarm Optimization.
• Incorporated Lévy mutation within the CNO to ensure diversity and prevent premature convergence.

Main Results:

• The proposed CNO method demonstrated effective global optimization for hash bit selection.
• Theoretical analysis confirmed the almost sure convergence of the CNO with Lévy mutation to global optima.
• Experimental results on three benchmarks showed the superiority of the CNO-based method over existing techniques.

Conclusions:

• The Collaborative Neurodynamic Optimization approach with Lévy mutation offers a robust and effective solution for hash bit selection.
• This method provides a significant advancement in optimizing hash bit subsets, outperforming conventional approaches.