Reference library
Evolving neural networks through augmenting topologies.
Paper summary
NEAT aligns connection genes by historical innovation identifiers, inherits matching genes from either parent and unmatched genes primarily from the fitter parent, with speciation and incremental structural growth. On double-pole balancing with velocity inputs, its crossover ablation averages 3,600 versus 5,557 evaluations across 120 runs per condition, p=0.001, with zero reported failures in these two conditions. The approximately 35.2% reduction is task/system-specific.
Relevance to the project
Assistant interpretation: historical homology is a useful comparison for matching inherited builder/type/controller genes before recombination. Gene ancestry does not guarantee continued functional equivalence or preserve all coupled F/H/updater semantics.
A specific older control benchmark within NEAT; not a universal crossover benefit or wall-clock comparison. Fixed genealogy cannot by itself resolve coadapted online-plastic modules.
Source record
- Reference ID
- doi-10-1162-106365602320169811
- Archived version
- Author-hosted published Evolutionary Computation PDF
- Summary provenance
- reference-library/papers/doi-10-1162-106365602320169811/metadata.json · summary
- Review status
- Targeted primary method/ablation review; earlier orientation record preserved in review_history
- Reading scope
- Targeted original PDF pp.9–11,13–14,19–22 on encoding, historical alignment, settings and component ablations, especially Table 3 and section 5.6; Figure4 caption checked through author-hosted paper text. Not exhaustive 30-page reading or independent reproduction.