BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260624T171341Z
LOCATION:Bldg. 8 - B 102
DTSTART;TZID=Europe/Stockholm:20260629T133000
DTEND;TZID=Europe/Stockholm:20260629T153000
UID:submissions.pasc-conference.org_PASC26_sess111@linklings.com
SUMMARY:MS1H - From Physics to Data - and Back: Trustworthy Machine Learni
 ng Potentials for Materials Design
DESCRIPTION:Organizer(s): Michał Sanocki, Ian Störmer, Julija Zavadlav Kol
 ler (Technical University of Munich), and Philip Robin Loche (EPFL)\n\nMac
 hine learning interatomic potentials (MLIPs) are transforming electronic-s
 tructure modeling by delivering near–quantum mechanical accuracy at greatl
 y reduced cost. As these models increasingly surrogate explicit electronic
 -structure calculations, especially for large, complex, or heterogeneous s
 ystems, the central challenge becomes establishing trust when direct quant
 um-mechanical validation is no longer feasible. Balancing data-driven flex
 ibility with physical consistency is essential: symmetry-preserving and ph
 ysics-informed architectures enhance reliability, while more flexible mode
 ls demand rigorous verification to ensure they learn the correct physical 
 relationships. Exascale computing reshapes this landscape by enabling larg
 e-scale data generation, systematic cross-validation, and interrogation of
  MLIP behavior under extreme conditions. At the same time, the complexity 
 of training and deploying these models highlights the need for reproducibi
 lity, explainability, and robust uncertainty quantification. Emerging appr
 oaches aim to embed trust checks - such as enforcing conservation laws, eq
 uivariance, and stability-directly into training rather than relying solel
 y on post hoc validation. This minisymposium will gather researchers from 
 chemistry, materials science, physics, and computer science to discuss str
 ategies for developing trusted MLIPs. Topics include physics-informed repr
 esentations, equivariant architectures, long-range and non-conservative in
 teractions, uncertainty estimation, and automated verification workflows. 
 The session aims to chart pathways toward reliable, scalable, and transpar
 ent MLIP frameworks for molecular and materials simulation.\n\nOpen Discus
 sion: Trustworthy Machine Learning Potentials - Challenges and Perspective
 s\n\nThis session will be an open discussion following the presentations i
 n the minisymposium. The goal is to create space for speakers and particip
 ants to exchange perspectives on common challenges in developing and apply
 ing machine learning interatomic potentials. Topics raised by the talks in
 clude dat...\n\n\nMichał Sanocki (Technical University of Munich)\n-------
 --------------\nFrom Dataset-Induced Biases to Uncertainty-Driven Learning
  of Interatomic Potentials\n\nMachine-learned interatomic potentials have 
 significantly advanced atomistic modeling by combining near first-principl
 es accuracy with the efficiency of classical potentials. However, their re
 liability in simulations strongly depends on the datasets used for trainin
 g, both for specialized and founda...\n\n\nViktor Zaverkin (Saarland Unive
 rsity, Leibniz Institute for New Materials)\n---------------------\nQuantu
 m (Bio) Molecular Simulations with Machine Learning Force Fields\n\nMachin
 e learning force fields (MLFFs) promise to bridge the gap between quantum-
 mechanical accuracy and the computational efficiency needed to simulate re
 alistic (bio)molecular systems [1]. Yet their predictive power is often li
 mited by the quality and coverage of training data, as well as by locali..
 .\n\n\nAdil Kabylda (University of Luxembourg)\n---------------------\nEnd
 -to-End Uncertainty Quantification for Atomistic Machine Learning\n\nMachi
 ne-learning interatomic potentials (MLIPs) have found widespread adoption 
 in atomistic simulation workflows. Their use, however, introduces statisti
 cal uncertainties from finite training data and approximation errors, as w
 ell as systematic discrepancies inherited from the underlying electronic-.
 ..\n\n\nMatthias Kellner (EPFL)\n\nDomain: Chemistry and Materials, Physic
 s, Computational Methods and Applied Mathematics\n\nSession Chairs: Michał
  Sanocki (Technical University of Munich); Ian Störmer (Technical Universi
 ty of Munich); and Philip Loche (Technical University of Munich, EPFL)
END:VEVENT
END:VCALENDAR