BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//7.2.3.1//EN TZID:Europe/Berlin X-WR-TIMEZONE:Europe/Berlin BEGIN:VEVENT UID:1-198@dmg-ev.de DTSTART;TZID=Europe/Berlin:20250605T150000 DTEND;TZID=Europe/Berlin:20250605T170000 DTSTAMP:20250422T110625Z URL:https://www.dmg-ev.de/veranstaltungen/online-seminarserie-kuenstliche- intelligenz-in-der-hydrometeorologie-2/ SUMMARY:Online-Seminarserie: Künstliche Intelligenz in der Hydrometeorolog ie DESCRIPTION:Chaopeng Shen\, Pennsylvania State University:\n\nDifferentiabl e parameter learning and differentiable modeling\nAbstract\nBig data and a rtificial intelligence (AI) methods are revolutionizing how knowledge is g ained and predictions are made for sustainability sciences and the global environment. However\, purely data-driven AI models often suffer from perf ormance penalties when applied to data-scarce regions or extreme regimes\, besides lacking interpretability. Here we show that differentiable models (a genre of physics-informed machine learning where gradients can be rapi dly computed via a range of methods\, allowing process-based equations to be seamlessly trained together with neural networks\, https://t.co/qyuAzY PA6Y) are well-suited to capture unseen extremes because they utilize phys ical principles like mass balances and first-order exchanges to restrict t he role of neural networks. They offer highly competitive performance and simultaneously provide physical process clarity. We demonstrate how the ne xt-generation differentiable national-scale hydrologic model\, multivariat e water quality and ecosystem models achieve state-of-the-art results and open new avenues of learning. To improve computational efficiency\, we fu rther consider training highly efficient surrogate models for these model s (mostly parameterized ordinary or partial differential equations) with t he correct sensitivity. Recent advances in surrogate modeling\, particular ly deep learning frameworks like the Fourier Neural Operator (FNO)\, have demonstrated significant efficiency in approximating solution paths. Howev er\, these approximations result in inaccurate solutions to inverse proble ms due to inaccurate sensitivities and are highly sensitive to concept dri ft. We propose Sensitivity-Constrained Fourier Neural Operators (SC-FNO\, presented in AI conference ICLR 2025\, main session). SC-FNO ensures accur acy in the solution paths\, inverse problems\, and sensitivity calculation s\, even under sparse training data or concept drift scenarios. Differenti able modeling together with sensitivity-constrained neural operators are p osed to drastically improve our simulation and learning capabilities for a wide range of engineering and geoscientific problems.\nBiographical Sketc h\nChaopeng Shen is Professor in Civil Engineering at The Pennsylvania St ate University. He received the Ph.D. degree in environmental engineering from Michigan State University\, East Lansing\, MI\, USA\, in 2009. His Ph D research focused on computational hydrology and he developed the hydrolo gic model Process-based Adaptive Watershed Simulator (PAWS)\, which was la ter coupled to the community land model to study the interactions between hydrology and ecosystem. His recent efforts focused on harnessing the big data and machine learning (ML) and physics-informed ML opportunities in ad vancing hydrologic predictions and understanding. As an early advocate for ML in geosciences\, he has written technical\, editorial\, review and col lective opinion papers on hydrologic deep learning to call to attention th e emerging opportunities for scientific advances. He currently promotes  differentiable modeling which seamlessly integrates neural networks and physics for knowledge discovery. CATEGORIES:FA Hydrometeorologie LOCATION:https://uni-bonn.zoom-x.de/j/67265546395?pwd=lAlQlpxXrFB45IFLnEEpw Mvbzlbibz.1 END:VEVENT BEGIN:VTIMEZONE TZID:Europe/Berlin X-LIC-LOCATION:Europe/Berlin BEGIN:DAYLIGHT DTSTART:20250330T030000 TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST END:DAYLIGHT END:VTIMEZONE END:VCALENDAR