NEPHILA: A Bio-Inspired Optimization Algorithm Based on Adaptive Interception Dynamics and Learning-Guided Search

Abstract This paper introduces the NEPHILA Optimization Algorithm (NOA), a bio-inspired stochastic optimization framework motivated by the adaptive interception behavior of orb-weaving spiders. Rather than reproducing a literal biological model, NOA translates these dynamics into a learning-guided search architecture integrating Brownian perturbations, exponentially decaying Lévy flights, fitness-based adaptive pruning, elitist guidance, reinforcement-guided operator control, and a non-parametric Gaussian radial basis function model for estimating spatial solution quality.The proposed framework was investigated through continuous and combinatorial optimization experiments. Six standard continuous benchmark functions were considered, with 30 independent runs per function and comparisons against PSO, ACOR, FA, BA, GWO, and Differential Evolution. The results revealed problem-dependent performance, with GWO obtaining the lowest mean error on five functions and PSO performing best on Schwefel.To investigate applicability to discrete optimization, a permutation-based NEPHILA variant incorporating adaptive operator selection, edge-guided candidate selection, and bounded 2-opt improvement was evaluated on synthetic and TSPLIB Traveling Salesman Problem instances. On a synthetic 100-city instance, NEPHILA achieved the lowest mean tour length among the evaluated methods. However, comparisons with a GA-2opt baseline on four TSPLIB instances showed that NEPHILA obtained a lower mean tour length on one instance, while GA-2opt performed better on the remaining three.Additional experiments demonstrated that increasing the evaluation budget from 15,030 to 60,000 reduced the mean tour length by approximately 29.69% on kroA100. Ablation results further revealed that simplified operator configurations could improve solution quality, although differences in local-search effort prevented attributing these gains exclusively to individual operators.Overall, the findings establish NEPHILA as a flexible learning-guided stochastic optimization framework whose effectiveness depends on the objective landscape, computational budget, and coordination of search mechanisms. Rather than demonstrating universal superiority, the study identifies the conditions and algorithmic components that influence performance and provides an empirical basis for further refinement. MSC Classification: 90C26 , 90C59 , 68T05 , 68W20

Authors

Publication Details

Journal
Research Square
Published
2026-11-12
DOI
https://doi.org/10.21203/rs.3.rs-11368464/v1
Primary Topic
Metaheuristic Optimization Algorithms Research
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

NEPHILA: A Bio-Inspired Optimization Algorithm Based on Adaptive Interception Dynamics and Learning-Guided Search

Gilberto Pérez Lechuga
Research Square
Metaheuristic Optimization Algorithms Research
preprint

NEPHILA: A Bio-Inspired Optimization Algorithm Based on Adaptive Interception Dynamics and Learning-Guided Search

Gilberto Pérez Lechuga
preprint en

Abstract

Abstract This paper introduces the NEPHILA Optimization Algorithm (NOA), a bio-inspired stochastic optimization framework motivated by the adaptive interception behavior of orb-weaving spiders. Rather than reproducing a literal biological model, NOA translates these dynamics into a learning-guided search architecture integrating Brownian perturbations, exponentially decaying Lévy flights, fitness-based adaptive pruning, elitist guidance, reinforcement-guided operator control, and a non-parametric Gaussian radial basis function model for estimating spatial solution quality.The proposed framework was investigated through continuous and combinatorial optimization experiments. Six standard continuous benchmark functions were considered, with 30 independent runs per function and comparisons against PSO, ACOR, FA, BA, GWO, and Differential Evolution. The results revealed problem-dependent performance, with GWO obtaining the lowest mean error on five functions and PSO performing best on Schwefel.To investigate applicability to discrete optimization, a permutation-based NEPHILA variant incorporating adaptive operator selection, edge-guided candidate selection, and bounded 2-opt improvement was evaluated on synthetic and TSPLIB Traveling Salesman Problem instances. On a synthetic 100-city instance, NEPHILA achieved the lowest mean tour length among the evaluated methods. However, comparisons with a GA-2opt baseline on four TSPLIB instances showed that NEPHILA obtained a lower mean tour length on one instance, while GA-2opt performed better on the remaining three.Additional experiments demonstrated that increasing the evaluation budget from 15,030 to 60,000 reduced the mean tour length by approximately 29.69% on kroA100. Ablation results further revealed that simplified operator configurations could improve solution quality, although differences in local-search effort prevented attributing these gains exclusively to individual operators.Overall, the findings establish NEPHILA as a flexible learning-guided stochastic optimization framework whose effectiveness depends on the objective landscape, computational budget, and coordination of search mechanisms. Rather than demonstrating universal superiority, the study identifies the conditions and algorithmic components that influence performance and provides an empirical basis for further refinement. MSC Classification: 90C26 , 90C59 , 68T05 , 68W20

Research Square
Tecnológico Nacional de México
Metaheuristic Optimization Algorithms Research
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon