I develop modern machine-learning-based techniques to solve complex computational problems in natural sciences.

I am currently a research and teaching assistant at Universität Hamburg, department of mathematics and the group of Controlled Molecule Imaging (CMI) at the Center for Free-Electron Laser Science CFEL, Deutsches Elektronen-Synchrotron DESY.

• Languages: Python, Julia, TeX
• Machine learning libraries: Jax, PyTorch, TensorFlow

• Active learning (AL) for constructing potential energy surfaces: AL
• Spectral learning (SL) for solving differential equations: SL

Publications in chronological order:

• V. Menden, Y. Saleh, A. Iske, Bounds on the Generalization Error in Active Learning, arXiv:2409.09078 (2024) [stat.ML, cs.LG] (2024).
• Y. Saleh, A. Iske, Inducing Riesz and Orthonormal Bases in $L^2$ via Composition Operators, arXiv:2406.18613 (2024) [math.FA, math.NA, cs.LG] (2024).
• T. Wenzel, Y. Saleh, A. Iske, Two-Layered and Deep Kernels as Data-Adapted Kernels, DEEPK 2024 International Workshop on Deep Learning and Kernel Machines (2024).
• Y. Saleh, A.F. Corral, A. Iske, J. Küpper, A. Yachmenev, Computing Excited States of Molecules using Normalizing Flows, arXiv:2308.16468 (2023) [cs.LG, phys.chem-phys] (2023).
• Y. Saleh, Spectral and Active Learning for Efficient and Computationally Scalable Quantum Molecular Dynamics, Dissertation, Universität Hamburg (2023).
• Y. Saleh, A. Iske, A. Yachmenev, J. Küpper, Augmenting Basis Sets by Normalizing Flows, Proc. Appl. Math. Mech. 23 (1), e202200239 (2023).
• Y. Saleh, V. Sanjay, A. Iske, A. Yachmenev, J. Küpper, Active Learning of Potential-Energy Surfaces of Weakly Bound Complexes with Regression-Tree Ensembles, J. Chem. Phys. 155, 144109 (2021).

In chronological order:

• Computing Excited States of Molecules using Normalizing Flows, High-Dimensional Quantum Dynamics Conference, Hamburg, Germany (poster, 2024).
• Spectral Learning for Solving Molecular Schrödinger Equations, PINN-PAD: Physics Informed Neural Networks in PADova, Padova, Italy (contributed talk, 2024).
• Augmenting Spectral Methods with Invertible Neural Networks and Application to Quantum Molecular Physics, ALGORITMY conference, Podbanské, High Tatra Mountains, Slovakia (contributed talk, 2024).
• Learning Basis Sets in $L^2$ and Application to Computing Excited States of Molecules, Workshop on Modern Methods for Differential Equations of Quantum Mechanics, Banff International Research Station, Banff, Canada (invited talk, 2024).
• Flow-Induced Bases and Application to Quantum Molecular Physics, Machine Learning in Engineering, TU-Hamburg (invited talk, 2023).
• Augmenting Bases with Normalizing Flows for Solving Schrödinger Equations, Deutsche Physikalische Gesellschaft (DPG) SAMOP Frühjahrstagung (invited talk, 2023).
• Spectral Learning for Solving the Schrödinger Equation for Molecules, Conference on Computational Methods in Applied Mathematics, TU Wien (contributed talk, 2022).
• Deep Spectral Methods for Solving Variational Problems Arising from Differential Equations, Gesellschaft für Angewandte Mathematik und Mechanik (GAMM) Jahrestagung, TU Aachen (contributed talk, 2022).
• Deep Spectral Methods for Solving Variational Problems Arising from Differential Equations, Hausdorff Center for Mathematics Workshop: Synergies Between Data Science and PDE Analysis, Universität Bonn (contributed talk, 2022).
• Hausdorff School: Foundational Methods in Machine Learning (2022).
• Deep Learning for Computing Spectra of Molecules, Opening Symposium of the Center for Data and Computing in Natural Science (CDCS) (poster, 2022).
• Spectral Learning for (Ro-)Vibrational Calculations of Weakly-Bound Molecules, Deutsche Physikalische Gesellschaft (DPG) SAMOP Frühjahrstagung (contributed talk, 2022).
• Deep Learning for Computing Rovibrational Spectra of Molecules, European CFEL and DESY Photon Science Users’ Meeting (poster, 2022).
• Helmholtz H3 Hackathon (2021).
• Applications of Machine Learning in Quantum Simulations of Hydrogen Bond Dynamics, Warsaw Summer School for Quantum Physics and Chemistry, University of Warsaw (poster, 2021).
• Applications of Machine Learning in Quantum Simulations of Hydrogen Bond Dynamics, Machine Learning for Quantum X (invited talk, online, 2021).
• Active Learning for Computing Potential-Energy Surfaces of Molecules, Bunsentagung (contributed talk, online, 2021).