Reproduce “Complex System Exploration with Interactive Human Guidance” 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
Bastien Morel, Clément Moulin-Frier, Pascal Barla
Researchers at Univ. Bordeaux and INRIA developed NearestRandomAxesBalanced (NRAB), an interactive method for exploring complex systems that integrates human guidance through explicit, system-agnostic constraints within an Intrinsically Motivated Goal Exploration Process. This approach efficiently discovers diverse spatio-temporal patterns within user-defined regions of interest while maintaining global pattern diversity, demonstrating up to a 20x improvement in pattern discovery efficiency for Gray-Scott systems compared to random sampling.
Key insights
- Integrating explicit, system-agnostic constraints provides an efficient and flexible mechanism for real-time human guidance in complex system exploration.
- A balanced exploration strategy, simultaneously targeting user-defined ROIs and broader pattern diversity, ensures both focused discovery and adaptability to evolving user interests.
- Navigating high-dimensional and non-specific behavior spaces can be effectively achieved through techniques like standardization and random dimension selection for nearest neighbor searches, reducing reliance on prior system knowledge.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.