Autonomous (AI-driven) materials science

Applied Physics Reviews, Journal Year: 2022, Volume and Issue: 9(3)

Published: Sept. 1, 2022

First Page

---

Data storage architectures to accelerate chemical discovery: data accessibility for individual laboratories and the community

Chemical Science, Journal Year: 2022, Volume and Issue: 13(46), P. 13646 - 13656

Published: Jan. 1, 2022

With the increasing emphasis on data sharing, reproducibility, and replicability, big-data analytics, machine learning, chemists must consider database management systems for their laboratory's storage, management, accessibility.

Language: Английский

---

How to Accelerate R&D and Optimize Experiment Planning with Machine Learning and Data Science

CHIMIA International Journal for Chemistry, Journal Year: 2023, Volume and Issue: 77(1/2), P. 7 - 7

Published: Feb. 22, 2023

Accelerating R&D is essential to address some of the challenges humanity currently facing, such as achieving global sustainability goals. Today’s Edisonian approach trial-and-error still prevalent in labs takes up two decades fundamental and applied research for new materials reach market. Turning around this situation calls strategies upgrade expedite innovation. By conducting smart experiment planning that data-driven guided by AI/ML, researchers can more efficiently search through complex - often constrained space possible experiments find or hit optima much faster than with current approaches. Moreover, digitized data management, will be able maximize utility their short long terms aid statistics, ML visualization tools. In what follows, we describe a framework lay out key technologies accelerate optimize

Language: Английский

---

Decoding structure-spectrum relationships with physically organized latent spaces

Physical Review Materials, Journal Year: 2023, Volume and Issue: 7(5)

Published: May 16, 2023

A semisupervised machine learning method for the discovery of structure-spectrum relationships is developed and then demonstrated using specific example interpreting x-ray absorption near-edge structure (XANES) spectra. This constructs a one-to-one mapping between individual descriptors spectral trends. Specifically, an adversarial autoencoder augmented with rank constraint (RankAAE). The RankAAE methodology produces continuous interpretable latent space, where each dimension can track descriptor. As part this process, model provides robust quantitative measure relationship by decoupling intertwined contributions from multiple structural characteristics. makes it ideal interpretation descriptors. capability procedure showcased considering five local database >50 000 simulated XANES spectra across eight first-row transition metal oxide families. resulting not only reproduce known trends in literature but also reveal unintuitive ones that are visually indiscernible large datasets. results suggest has great potential to assist researchers complex scientific data, testing physical hypotheses, revealing patterns extend insight.

Language: Английский

---

Olympus, enhanced: benchmarking mixed-parameter and multi-objective optimization in chemistry and materials science

Published: May 18, 2023

Experiment planning algorithms are a required component of autonomous platforms for scientific discovery. Selecting suitable optimization algorithm novel application is an important yet difficult choice researcher has to make based on past empirical performance similar tasks. To facilitate the evaluation various chemistry and materials science tasks, we previously introduced OLYMPUS (Mach. Learn.: Sci. Technol. 2, 035021, 2021), Python package providing consistent easy-to-use interface numerous benchmark datasets. While original was limited continuous parameters single objectives, in this work expand OLYMPUS' capabilities include mixed (continuous, discrete, categorical) parameter types multiple objectives. Several experiment already contained extended handle categorical discrete types, five additional planners implemented (23 total). We also provide 23 datasets taken from literature (33 total), covering wide range research areas, chemical reaction manufacturing. Finally, visualization enhanced allow easy inspection results, core functionality embedded Streamlit web code-free usage. demonstrate how enables researchers rapidly different strategies gain insight into their behavior by focusing two case studies: Suzuki-Miyaura cross-coupling with conditions, multi-objective redox-active materials. The updated provides practitioners large suite tools efficiently analyze mixed-parameter

Language: Английский

---