Human in the loop automated experiments in Scanning probe microscope

We develop a human-AI collaborated workflow towards autonomous and aligned experiments, as we called it "Bayesian optimized active recommender system" which in short BOARS.
Like

Share this post

Choose a social network to share with, or copy the URL to share elsewhere

This is a representation of how your post may appear on social media. The actual post will vary between social networks

I am very glad to share our latest publication in npj Computational Materials of developing human in the loop Bayesian optimization workflow or we named it — Bayesian optimized spectral recommender system (BOARS) to allow the microscope operator/domain experts to custom design any material target properties on the fly (without any need of priory set up) they want to autonomously study structure-property relationships.

The motivation behind this tool is to improve AI alignment with domain scientist, particularly in the field of material discovery when the operator could find some interesting experiments in the course of autonomous exploration and want to steer the exploration accordingly. This level of experimental steering in real time is missing in classical Bayesian optimization, which resulted in increasing complexity of the microscope operators due to the need to stop the experiment and translate newly found interesting experimental data into an appropriate target function. However, the operator would like to have simpler operation where they have the flexibility to change the target or learning objective on the fly as they initially explore over the large unknown material image space. This is what exactly BOARS do, where in the early exploration human guides AI to find the desirable target property and once human is satisfied, the AI takes over to converge quickly to the optimal structure-human identified target property learning.

This can be easily adapted for other domain problems as well.

This work builds on large collaboration among researchers from Oak Ridge National Laboratory, University of Tennessee, Knoxville, Pacific Northwest National Laboratory, National Cheng Kung University.

This work has a special place in my heart firstly because as an ML expertise, I had a first time hands on tutorial of how scanning probe microscope work.


For more details, please follow https://doi.org/10.1038/s41524-023-01191-5


Now with my future plans, more developments to follow to align AI with domain scientist in the field of autonomous systems.

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Follow the Topic

Scanning Probe Microscopy
Physical Sciences > Materials Science > Materials Characterization Technique > Microscopy > Scanning Probe Microscopy
Machine Learning
Mathematics and Computing > Computer Science > Artificial Intelligence > Machine Learning
Optimization
Mathematics and Computing > Mathematics > Optimization
Materials Characterization Technique
Physical Sciences > Materials Science > Materials Characterization Technique

Related Collections

With collections, you can get published faster and increase your visibility.

Machine Learning Interatomic Potentials in Computational Materials

Publishing Model: Open Access

Deadline: Jun 06, 2025

Self-Driving Laboratories for Chemistry and Materials Science

Publishing Model: Open Access

Deadline: Jul 08, 2025