Choose timezone
Your profile timezone:
Explainable AI models for Materials Science with the SISSO Approach
—
Europe/Berlin
Zoom: https://us06web.zoom.us/j/84333961258
Zoom: https://us06web.zoom.us/j/84333961258
,
Save the date
(Calendar Entry / Zoom Link)
(Calendar Entry / Zoom Link)
Join us for this webinar, consisting of expert talks and interactive hands-on sessions, on the Sure Independence Screening and Sparsifying Operator (SISSO) approach for explainable artificial intelligence in materials science.
In an era where artificial intelligence is revolutionizing materials discovery, the challenge of creating interpretable and physically meaningful models remains paramount. Traditional AI approaches often produce "black box" models that, while accurate, provide little insight into the underlying physics governing material properties. The SISSO method addresses this gap by combining symbolic regression with compressed sensing to identify nonlinear analytical expressions that relate materials properties to key physicochemical descriptors.
SISSO has emerged as a particularly powerful tool for materials science applications where datasets are typically small, but the underlying physics is complex. Unlike conventional machine learning approaches that require extensive training data, SISSO can identify meaningful structure-property relationships from limited experimental or computational datasets. This deterministic approach generates interpretable models in the form of analytical expressions, enabling researchers to understand which physical parameters are most critical for a given property and how they interrelate.
The webinar will explore both the theoretical foundations and practical applications of SISSO, with particular emphasis on recent methodological advances and their implementation in catalysis research. Thomas Purcell will introduce the fundamental concepts of the SISSO method and discuss recent advances. These include improved feature representation through binary expression trees, enhanced treatment of physical units, parametric SISSO for greater flexibility, and refined solver algorithms for both regression and classification problems.
Building on these foundations, Lucas Foppa will introduce and demonstrate the advantages of hierarchical SISSO, showcasing real-world applications in heterogeneous catalysis. He will illustrate how SISSO can be applied to experimental data and combined with theoretical data to identify the "materials genes" of catalytic systems—those key physicochemical parameters that correlate with catalytic performance. Through concrete examples from alkane selective oxidation and other catalytic processes, the presentation will show how SISSO enables the identification of design rules that can guide the development of improved catalysts.
The accompanying online tutorial makes these advanced techniques accessible to both computational experts and experimentalists who want to leverage AI for materials design.
What will be covered?
-
Discover the fundamental principles and recent advances of the SISSO method, including algorithmic improvements in SISSO++ and hierarchical approaches
-
Explore practical applications of SISSO in catalysis research, from experimental data analysis to identifying structure-property relationships and design principles
-
Participate in a guided hands-on tutorial using our web-based application with an accessible, user-friendly interface that combines active learning, SISSO, and FHI-aims calculations
Both the webinar and hands-on session are free to attend without registration. The hands-on session addresses novice users of FHI-aims and SISSO, but may also be interesting to experienced users. Following along and running the calculations in the AWS cloud is completely free, allowing you to gain practical experience with FHI-aims and SISSO.
We provide resources for running the calculations during the hands-on session, but you need to register in advance to obtain access to the compute resources.
Register now!
Speakers:
Talk 1: An Introduction to SISSO and Recent Advances in the Methodology
Thomas A. R. Purcell
Assistant Professor, The University of Arizona, USA
Talk 2: Describing Materials Properties and Functions via the "Materials Genes" Concept
Lucas Foppa
Group Leader, NOMAD Laboratory at FHI of the Max Planck Society, Germany
When?
|
Webinar Talks + Q&A Wednesday, 03.09.2025: 09:30 - 11:00 EDT (USA East) 15:30 - 17:00 CEST (Europe) 19:00 - 20:30 IST (India)
|
Hands-On and Discussions Session 1 Thursday, 04.09.2025: 09:00 - 10:30 CEST (Europe) 12:30 - 14:00 IST (India) 15:00 - 16:30 CST (China) 16:00 - 17:30 JST (Japan) |
Hands-On and Discussions Session 2 Thursday, 04.09.2025: 07:00 - 08:30 PDT (USA Pacific) 10:00 - 11:30 EDT (USA East) 16:00 - 17:30 CEST (Europe)
|
Please choose the hands-on and discussion block that fits your schedule the best.
The hands-on session will assume a basic understanding of the FHI-aims input files. If you don't have experience with FHI-aims, you can learn the first steps from this short YouTube video:
The Basics of FHI-aims
Registration
Registration
Contact
