A new research paper revisits Lamarckian and Baldwinian evolution within evolutionary algorithms (EAs), comparing them against Darwinian evolution. Empirical results across six datasets for Maximum Independent Set and Maximum Cut problems demonstrate that Baldwinian and Lamarckian EAs consistently outperform Darwinian EAs, and often surpass recent deep learning baselines. Theoretical analysis also suggests Baldwinian evolution is asymptotically faster than Lamarckian, which is faster than Darwinian evolution for block lengths greater than two. AI
IMPACT This research suggests evolutionary algorithms, particularly Baldwinian and Lamarckian approaches, may offer competitive or superior performance to deep learning methods on certain graph-based problems.
RANK_REASON The cluster contains a research paper published on arXiv detailing empirical and theoretical comparisons of evolutionary algorithms.
Read on arXiv cs.NE (Neural & Evolutionary) →
- Baldwinian evolution
- deep learning
- evolutionary algorithms
- GraphBench
- Maximum Cut
- Maximum Independent Set
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
