Reference library
Parameterless Gene-pool Optimal Mixing Evolutionary Algorithms
Paper summary
GOMEA learns a family of variable subsets, commonly a mutual-information linkage tree, and copies donor values for a subset before keeping or reverting the change according to observed fitness. Subsets can be nested, overlapping and noncontiguous in storage. CGOM restricts donors using dependent variables outside the current subset. Forced improvement uses elitist material and can eventually copy an elite solution. In nine binary optimization benchmarks, CGOMEA-P3 improves on GOMEA-P3 on five, four significantly, but performs worse on HIFF. Evaluation efficiency does not uniformly imply faster runtime; population-size automation leaves other choices, including a tuned threshold.
Relevance to the project
Assistant interpretation: learn groups to exchange rather than assume a scalar cut in an array. Dependency estimation and incremental acceptance are candidates only after defining an inherited genome and a stable evaluation; noisy ecological outcomes require separate controls.
Binary combinatorial benchmarks do not establish effectiveness for recurrent plastic neural controllers or open-ended ecology. False or missed dependencies, estimation cost, noisy fitness, diversity loss and greedy rejection of temporarily worse steps remain concerns.
Source record
- Reference ID
- parameterless-gomea-2021
- Archived version
- arXiv:2109.05259v1, 11 September 2021
- Summary provenance
- reference-library/papers/parameterless-gomea-2021/metadata.json · summary
- Review status
- Targeted primary full-text review; not exhaustive reading or reproduction
- Reading scope
- Title/abstract identity and targeted PDF pp.3–6 (linkage, GOM, forced improvement and conditional mixing), p.8 (selected benchmarks), pp.12–13 (comparison results and runtime caveats). Other benchmark definitions, full tuning, supplement and figures not comprehensively audited; no reproduction.