Skip to content
Technical University of Munich

News

New work on "Efficient Bayesian multi-fidelity inverse analysis for expensive and non-differentiable physics-based simulations in high stochastic dimensions"

Abstract:High-dimensional Bayesian inverse analysis (dim >> 100) is mostly unfeasible for computationally demanding, nonlinear physics-based high-fidelity (HF) models. Usually, the use of more efficient gradient-based inference schemes is impeded if the multi-physics models are provided by complex…

New work on "Energy-Based Coarse-Graining in Molecular Dynamics: A Flow-Based Framework Without Data"

Abstract: Coarse-grained (CG) models offer an effective route to reducing the complexity of molecular simulations, yet conventional approaches depend heavily on long all-atom molecular dynamics (MD) trajectories to adequately sample configurational space. This data-driven dependence limits their…

Ece Sankar presents her Interdisciplinary Project on "Inverse Design for Molecules"

Abstract:  The goal of this interdisciplinary project is to enhance baseline models and methods for generative molecular modeling, aiming to generate novel molecules and perform inverse design to obtain molecular structures with desired properties. This project builds upon a previous…

New work on "DGenNO: A Novel Physics-aware Neural Operator for Solving Forward and Inverse PDE Problems based on Deep, Generative Probabilistic Modeling"

Abstract:  Solving parametric partial differential equations (PDEs) and associated PDE-based, inverse problems is a central task in engineering and physics, yet existing neural operator methods struggle with high-dimensional, discontinuous inputs and require large amounts of {\em labeled} training…

New paper in Machine Learning Science & Technology on "Stochastic Black-Box Optimization using Multi-Fidelity Score Function Estimator"

Optimizing parameters of physics-based simulators is crucial in the design process of engineering and scientific systems. This becomes particularly challenging when the simulator is stochastic, computationally expensive, black-box and when a high-dimensional vector of parameters needs to be…

New paper in Acta Materialia on "PSP-GEN: Stochastic inversion of the Process–Structure–Property chain in materials design through deep, generative probabilistic modeling"

Highlights • PSP-GEN integrates the entire PSP chain into one deep model, addressing complexities. • It employs a two-part latent space: one for microstructures and one for processing links. • Discrete-valued design is reformulated as continuous for gradient-based optimization. • PSP-GEN…

New paper in CMAME on "Weak neural variational inference for solving Bayesian inverse problems without forward models: Applications in elastography"

Highlights • Novel Bayesian method using ML and physics to solve model-based, inverse problems. • Uses weighted residuals as virtual observables, avoiding a classical forward model. • Solving inverse problems for ill-posed (insufficient bound. cond.) forward problems. • Demonstrations in…

Radu Vasilache presents his B.Sc thesis on "Material Behavior Prediction using Physics-Informed Neural Networks and Fourier Neural Operators"

New work on "PSP-GEN: Stochastic inversion of the Process-Structure-Property chain in materials design through deep, generative probabilistic modeling"

A framework for the stochastic inversion of Process-Structure-Property chain in the design of heterogeneous materials. More details: https://arxiv.org/abs/2408.01114

New work on "Weak neural variational inference for solving Bayesian inverse problems without forward models: applications in elastography"

We solve model-based Bayesian inverse problems without a forward model.We explore applications in elastography with the ultimate goal of enabling model-based diagnosis on hand-held devices by non-experts. More details here: https://arxiv.org/abs/2407.20697

To top