Machine learning enabling high-throughput and remote operations at large-scale user facilities

Abstract

Imaging, scattering, and spectroscopy are fundamental in understanding and discovering new functional materials. Contemporary innovations in automation and experimental techniques have led to these measurements being performed much faster and...

Linking scientific instruments and computation: Patterns, technologies, and experiences

“…In this increasingly popular approach to data reduction, previously collected data (from current or prior experiments and/or simulations) are used to train ML models to recognize interesting phenomena for data reduction or rapid response. 26 , 27 , 28 , 29 , 30 …”

Self-driving Multimodal Studies at User Facilities

“…The agents designed for this experiment considered what was rational, time sensitive, and interesting. The agents were developed using an ask-tell-report grammar, where an agent could be 'told' about new data, 'asked' what to do next, and 'report' about it's current status [5]. Each step was recorded using the streaming event model of Bluesky, via Tiled, so that the agent's perspective and decision making could be played back for investigation.…”

“…Due to advances in both light source accelerator and detector technologies, the productivity of a high-throughput beamline is no longer limited by the amount of photons it can produce and detect, but rather the ability to control and analyze the high rate of measurements. To help realize and leverage the full potential of these facilities, researchers have automated data collection and integrated artificial intelligence (AI) into the real-time analysis and orchestration of experiments [4,5,6]. In line with this, recent advancements have been made to convert and incorporate beamlines into self-driving labs or materials acceleration platforms (MAPs) [7,8,9,10].…”

When not to use machine learning: A perspective on potential and limitations

“…Cruising behind the slipstream created by tremendous success in the technology sector, ML has found wide applicability in the materials, chemical, and physical sciences. For example, the discovery, characterization and design of new materials, molecules and nanoparticles, [13][14][15][16][17][18][19][20][21][22][23][24] surrogate models for spectroscopy and other properties, [25][26][27][28] self-driving laboratories/autonomous experimentation, [29][30][31][32][33][34] and neural network potentials [35][36][37][38][39] have all been powered by ML and related methods. The current state of ML in materials science specifically has also been thoroughly documented in many excellent reviews 15,[40][41][42] that cover subject matter ranging from applications to computational screening and interpretation.…”