Reproduce “Exploring Promising Stepping Stones by Combining Novelty Search with Interactive Evolution” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Brian G. Woolley, Kenneth O. Stanley
Novelty-Assisted Interactive Evolutionary Computation (NA-IEC) combines human guidance, automated novelty search, and objective-based optimization to navigate deceptive search spaces. This approach discovers solutions in deceptive maze navigation tasks up to 9.1 times faster in clock time and 8 times faster in evaluations, with significantly lower genomic complexity, compared to fully automated methods, including a non-deceptive baseline.
Key insights
- Intermittent human intuition can effectively identify 'promising stepping stones' and guide the evolutionary search without requiring a formal, explicit, and potentially deceptive objective function.
- Strategically combining human guidance with automated novelty search dramatically improves search efficiency and reliability, enabling faster and more consistent discovery in deceptive domains.
- Automating the extensive exploration process via novelty search significantly reduces human fatigue, making interactive evolution viable for longer and more complex problem-solving sessions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.