A critical examination of robustness and generalizability of machine learning prediction of materials properties

npj Computational Materials, Journal Year: 2023, Volume and Issue: 9(1)

Published: April 7, 2023

Abstract Recent advances in machine learning (ML) have led to substantial performance improvement material database benchmarks, but an excellent benchmark score may not imply good generalization performance. Here we show that ML models trained on Materials Project 2018 can severely degraded new compounds 2021 due the distribution shift. We discuss how foresee issue with a few simple tools. Firstly, uniform manifold approximation and projection (UMAP) be used investigate relation between training test data within feature space. Secondly, disagreement multiple illuminate out-of-distribution samples. demonstrate UMAP-guided query by committee acquisition strategies greatly improve prediction accuracy adding only 1% of data. believe this work provides valuable insights for building databases enable better robustness generalizability.

Language: Английский

---

Exploiting redundancy in large materials datasets for efficient machine learning with less data

Nature Communications, Journal Year: 2023, Volume and Issue: 14(1)

Published: Nov. 10, 2023

Extensive efforts to gather materials data have largely overlooked potential redundancy. In this study, we present evidence of a significant degree redundancy across multiple large datasets for various material properties, by revealing that up 95% can be safely removed from machine learning training with little impact on in-distribution prediction performance. The redundant is related over-represented types and does not mitigate the severe performance degradation out-of-distribution samples. addition, show uncertainty-based active algorithms construct much smaller but equally informative datasets. We discuss effectiveness in improving robustness provide insights into efficient acquisition training. This work challenges "bigger better" mentality calls attention information richness rather than narrow emphasis volume.

Language: Английский

---

Recent advances in artificial intelligence boosting materials design for electrochemical energy storage

Chemical Engineering Journal, Journal Year: 2024, Volume and Issue: 490, P. 151625 - 151625

Published: April 24, 2024

In the rapidly evolving landscape of electrochemical energy storage (EES), advent artificial intelligence (AI) has emerged as a keystone for innovation in material design, propelling forward design and discovery batteries, fuel cells, supercapacitors, many other functional materials. This review paper elucidates burgeoning role AI materials from foundational machine learning (ML) techniques to its current pivotal advancing frontiers science storage, including enhancing performance, durability, safety battery technologies, cell efficiency longevity, fine-tuning supercapacitors achieve superior capabilities. Collectively, we present comprehensive overview recent advancements that have significantly accelerated development next-generation EES, offering insights into future research trajectories potential unlock new horizons science.

Language: Английский

---

Review of low-cost self-driving laboratories in chemistry and materials science: the “frugal twin” concept

Digital Discovery, Journal Year: 2024, Volume and Issue: 3(5), P. 842 - 868

Published: Jan. 1, 2024

Low-cost self-driving labs (SDLs) offer faster prototyping, low-risk hands-on experience, and a test bed for sophisticated experimental planning software which helps us develop state-of-the-art SDLs.

Language: Английский

---

Perspective: Machine learning in experimental solid mechanics

Journal of the Mechanics and Physics of Solids, Journal Year: 2023, Volume and Issue: 173, P. 105231 - 105231

Published: Jan. 31, 2023

Language: Английский

---

What is missing in autonomous discovery: open challenges for the community

Digital Discovery, Journal Year: 2023, Volume and Issue: 2(6), P. 1644 - 1659

Published: Jan. 1, 2023

Self-driving labs (SDLs) leverage combinations of artificial intelligence, automation, and advanced computing to accelerate scientific discovery.

Language: Английский

---

The future of self-driving laboratories: from human in the loop interactive AI to gamification

Digital Discovery, Journal Year: 2024, Volume and Issue: 3(4), P. 621 - 636

Published: Jan. 1, 2024

Self-driving laboratories (SDLs) are the future for scientific discovery in a world growing with artificial intelligence. The interaction between scientists and automated instrumentation leading conversations about impact of SDLs on research.

Language: Английский

---

Machine Learning Techniques for Improving Nanosensors in Agroenvironmental Applications

Agronomy, Journal Year: 2024, Volume and Issue: 14(2), P. 341 - 341

Published: Feb. 7, 2024

Nanotechnology, nanosensors in particular, has increasingly drawn researchers’ attention recent years since it been shown to be a powerful tool for several fields like mining, robotics, medicine and agriculture amongst others. Challenges ahead, such as food availability, climate change sustainability, have promoted pushed forward the use of agroindustry environmental applications. However, issues with noise confounding signals make these tools non-trivial technical challenge. Great advances artificial intelligence, more particularly machine learning, provided new that allowed researchers improve quality functionality nanosensor systems. This short review presents latest work analysis data from using learning agroenvironmental It consists an introduction topics application field nanosensors. The rest paper examples techniques utilisation electrochemical, luminescent, SERS colourimetric classes. final section discussion conclusion concerning relevance material discussed future sector.

Language: Английский

---

AI for Technoscientific Discovery: A Human-Inspired Architecture

Journal of Creativity, Journal Year: 2024, Volume and Issue: 34(2), P. 100077 - 100077

Published: Feb. 8, 2024

We present a high-level architecture for how artificial intelligences might advance and accumulate scientific technological knowledge, inspired by emerging perspectives on human such knowledge. Agents knowledge exercising technoscientific method—an interacting combination of engineering methods. The method maximizes quantity we call "useful learning" via more-creative implausible utility (including the "aha!" moments discovery), as well less-creative plausible utility. Society accumulates advanced agents so that other can incorporate build to make further advances. proposed is challenging but potentially complete: its execution in principle enable an equivalent full range

Language: Английский

---

Integrating autonomy into automated research platforms

Digital Discovery, Journal Year: 2023, Volume and Issue: 2(5), P. 1259 - 1268

Published: Jan. 1, 2023

The strict specification required for automatization to efficiently and reproducibly act in familiar domains restricts the flexibility needed autonomy when exploring new domains, requiring self-driving labs balance automation.

Language: Английский

---