Experimental Results for Hybrid PSO-DE Algorithms on 24 COCO/BBOB Functions

800 different hybridizations of Particle Swarm Optimization and Differential Evolution algorithms were benchmarked on a set of 24 continuous black-box test functions. The data includes expected running time metrics and raw runtimes from 30 independent runs per algorithm-problem pair, generated for a GECCO '20 conference paper by researchers from Leiden University.

Use Cases

• Comparing the performance of different hybrid PSO-DE algorithm configurations based on expected running time (ERT) metrics.
• Analyzing the variability of algorithm performance across multiple independent runs using the provided standard deviation values.
• Benchmarking new optimization algorithms against the 800 hybrid variants on the COCO/BBOB problem set.
• Reproducing the empirical cumulative distribution function (ECDF) plots from the original study using the provided R scripts.

Strengths

• Benchmarks 800 distinct algorithm hybridizations, providing a broad comparison space.
• Includes results from 30 independent runs per algorithm-problem pair, allowing for statistical analysis.
• Data is associated with a peer-reviewed conference publication (GECCO '20), suggesting a level of methodological rigor.

Limitations

• Column-level documentation is absent; field semantics must be inferred from the description or R scripts after download.
• Row count is unknown, which may limit suitability assessment for large-scale analyses.
• Last update date is unknown; freshness unverified.

Provenance

Source

Leiden University

Collection Method

Computational experiments benchmarking algorithm variants on the COCO/BBOB problem set.