BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:FAIRmat Seminar with Abhishek K. Singh
DTSTART:20261005T120000Z
DTEND:20261005T130000Z
DTSTAMP:20260925T222200Z
UID:indico-event-55@events.fairmat-nfdi.eu
CONTACT:fairmat-events@physik.hu-berlin.de
DESCRIPTION:Interpretable Representations in Materials AIThe FAIRmat Semin
 ar series continues with a talk by Prof. Abhishek K. Singh\, Chair at t
 he Materials Research Centre\, Indian Institute of Science.Abstract:Mate
 rials representation across multiple length scales is essential for enabli
 ng AI models to solve structure-to-property prediction problems in complex
  systems such as superalloys. Properties like Vickers hardness are primari
 ly governed by microstructural features\, while formation energy correlat
 es to atomic arrangements\, making it crucial to capture relevant informat
 ion from different structural hierarchies. Accurate and interpretable repr
 esentations across these scales allow machine learning approaches to accel
 erate materials discovery and design. At the microstructural level\, thre
 e frameworks are used to represent image-based information: (1) statistica
 l representations using 2-point spatial correlations to capture phase dist
 ribution patterns\, (2) geometry-driven image processing techniques that e
 xtract morphological descriptors such as area\, perimeter\, and shape of p
 recipitates\, and (3) deep learning models like convolutional neural netwo
 rks that automatically learn hierarchical features directly from raw SEM i
 mages. These image-derived features are combined with metadata such as com
 position and processing history to predict mechanical properties such as 
 Vickers hardness. At the atomic level\, graph-based representations like 
 the CLEAR (Chemistry and Local Environment Adaptive Representation) descri
 ptor-based model represent crystal structures by combining elemental prope
 rties with interatomic distances through Voronoi-based neighbours. By appl
 ying pooling operations\, these graph features are transformed into fixed-
 size vectors that enable predictive modelling of formation energy and phas
 e stability.1. Acta Mater.\, 196\, 295-303 (2020)2. J. Mater. Sci. 55\, 15
 845 (2020)3. Acta Mater.\, 276\, 120-122 (2024)4. Comput. Mater. Sci (2025
 )\, 113854  Join us for this engaging discussion in person at the Center
  for the Science of Materials Berlin or online. No registration is requir
 ed for in-person attendance. To receive the Zoom link for online participa
 tion\, please complete the registration form.\n\nhttps://events.fairmat-nf
 di.eu/event/55/
IMAGE;VALUE=URI:https://events.fairmat-nfdi.eu/event/55/logo-1943420652.pn
 g
LOCATION:1.107 (Center for the Science of Materials Berlin (CSMB))
URL:https://events.fairmat-nfdi.eu/event/55/
END:VEVENT
END:VCALENDAR