Research map / Probabilistic and scientific modeling / Molecules and proteins
Adaptive Experimental Design to Accelerate Scientific Discovery and Engineering Design
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
An AAAI 2025 talk abstract summarizing the author's research on adaptive experimental design (Bayesian optimization) for scientific discovery and engineering design, covering dictionary-embedding GP models and information-theoretic multi-fidelity multi-objective policies.
Problem
Efficiently searching large combinatorial or high-dimensional design spaces when each candidate evaluation requires an expensive lab experiment or simulation.
Approach
Summarizes prior work: dictionary-embedding Gaussian process models for high-dimensional binary/categorical spaces with regret bounds tied to dictionary orthogonality, LADDER kernels combining deep generative model latent spaces with structured kernels, and output space entropy based information-per-cost policies for multi-fidelity multi-objective optimization.
Key results
- Dictionary-embedding GP bandit regret bounds depend on dictionary orthogonality, linking GP modeling of combinatorial structures to compressed sensing.
- Output space entropy policies significantly improved state of the art in multi-fidelity multi-objective design optimization.
- Methods identified optimal covalent organic frameworks for methane storage and noble gas separation while evaluating only a small fraction of candidates.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Combine latent-space kernels from deep generative models with expert-designed structured kernels (LADDER) | LADDER · Molecule design benchmarks | optimization performance | Helped |
| Use output space entropy-based information-theoretic decision-making policies | Multi-fidelity multi-objective design optimization · Multi-fidelity multi-objective design optimization problems | optimization performance | Helped |
Limitations
- Abstract-level summary without experimental details or quantitative results.
Experimental setup
Builds on: Gaussian process bandits LADDER output space entropy
Closest papers by meaning
- EngiBench: A Framework for Data-Driven Engineering Design Research (NeurIPS 2025 · similarity 0.84)
- AI-Researcher: Autonomous Scientific Innovation (NeurIPS 2025 · similarity 0.84)
- Quality-Weighted Vendi Scores And Their Application To Diverse Experimental Design (ICML 2024 · similarity 0.83)
- AutoSciLab: A Self-Driving Laboratory for Interpretable Scientific Discovery (AAAI 2025 · similarity 0.83)
- LLM and Simulation as Bilevel Optimizers: A New Paradigm to Advance Physical Scientific Discovery (ICML 2024 · similarity 0.83)
- PASOA- PArticle baSed Bayesian Optimal Adaptive design (ICML 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.