Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

How Useful is Intermittent, Asynchronous Expert Feedback for Bayesian Optimization?

Agustinus Kristiadi, Felix Strieth-Kalthoff, Sriram Ganapathi Subramanian, Vincent Fortuin, Pascal Poupart, Geoff Pleiss

TL;DR

This work investigates whether a small amount of randomly arriving, non-blocking expert feedback can meaningfully steer Bayesian optimization within self-driving laboratories. By training a Bayesian preference model on asynchronous feedback from a separate thread and integrating it into the BO loop via a preference-aware acquisition, the approach aims to retain automation while leveraging human intuition. Across toy and chemistry datasets, even 10–25% intermittent feedback can improve data efficiency and steer exploration toward preferred criteria, though careful tuning (e.g., annealing the reliance on expert bias) is important. The results suggest a viable pathway to data-efficient, human-guided autonomous discovery, with avenues for improved active-learning and real-world evaluation.

Abstract

Bayesian optimization (BO) is an integral part of automated scientific discovery -- the so-called self-driving lab -- where human inputs are ideally minimal or at least non-blocking. However, scientists often have strong intuition, and thus human feedback is still useful. Nevertheless, prior works in enhancing BO with expert feedback, such as by incorporating it in an offline or online but blocking (arrives at each BO iteration) manner, are incompatible with the spirit of self-driving labs. In this work, we study whether a small amount of randomly arriving expert feedback that is being incorporated in a non-blocking manner can improve a BO campaign. To this end, we run an additional, independent computing thread on top of the BO loop to handle the feedback-gathering process. The gathered feedback is used to learn a Bayesian preference model that can readily be incorporated into the BO thread, to steer its exploration-exploitation process. Experiments on toy and chemistry datasets suggest that even just a few intermittent, asynchronous expert feedback can be useful for improving or constraining BO. This can especially be useful for its implication in improving self-driving labs, e.g. making them more data-efficient and less costly.

How Useful is Intermittent, Asynchronous Expert Feedback for Bayesian Optimization?

TL;DR

This work investigates whether a small amount of randomly arriving, non-blocking expert feedback can meaningfully steer Bayesian optimization within self-driving laboratories. By training a Bayesian preference model on asynchronous feedback from a separate thread and integrating it into the BO loop via a preference-aware acquisition, the approach aims to retain automation while leveraging human intuition. Across toy and chemistry datasets, even 10–25% intermittent feedback can improve data efficiency and steer exploration toward preferred criteria, though careful tuning (e.g., annealing the reliance on expert bias) is important. The results suggest a viable pathway to data-efficient, human-guided autonomous discovery, with avenues for improved active-learning and real-world evaluation.

Abstract

Bayesian optimization (BO) is an integral part of automated scientific discovery -- the so-called self-driving lab -- where human inputs are ideally minimal or at least non-blocking. However, scientists often have strong intuition, and thus human feedback is still useful. Nevertheless, prior works in enhancing BO with expert feedback, such as by incorporating it in an offline or online but blocking (arrives at each BO iteration) manner, are incompatible with the spirit of self-driving labs. In this work, we study whether a small amount of randomly arriving expert feedback that is being incorporated in a non-blocking manner can improve a BO campaign. To this end, we run an additional, independent computing thread on top of the BO loop to handle the feedback-gathering process. The gathered feedback is used to learn a Bayesian preference model that can readily be incorporated into the BO thread, to steer its exploration-exploitation process. Experiments on toy and chemistry datasets suggest that even just a few intermittent, asynchronous expert feedback can be useful for improving or constraining BO. This can especially be useful for its implication in improving self-driving labs, e.g. making them more data-efficient and less costly.
Paper Structure (13 sections, 2 equations, 11 figures)

This paper contains 13 sections, 2 equations, 11 figures.

Table of Contents

  1. Introduction
  2. Background
  3. Bayesian optimization
  4. Preference learning
  5. Laplace approximations
  6. Experiments
  7. Setups
  8. Results
  9. Conclusion
  10. Related Work
  11. Additional results
  12. On Thompson sampling
  13. Active Learning

Figures (11)

  • Figure 1: Accelerated scientific discovery takes humans out of the loop (black). We investigate the usefulness of human feedback that is incorporated in a compatible way to that spirit (red).
  • Figure 2: Bayesian Optimization
  • Figure 3: Preference Learning
  • Figure 5: Results with random selection on the expert-feedback thread. Dashed black lines are the optimum values.
  • Figure 6: Toy BO results, simulating expert feedback on $x$ that is conflicting with $f(x)$. The dotted curve represents the norm of the current best $x$.
  • ...and 6 more figures