60 https://www.uni-bremen.de/techmath/veranstaltungen/2026-komso-academy Ankündigung und Programm der KoMSO Academy de Universität Bremen Wed, 20 May 2026 04:08:32 +0200 Wed, 20 May 2026 04:08:32 +0200 Universität Bremen content-722404 Mon, 04 May 2026 10:58:07 +0200 https://www.uni-bremen.de/techmath/veranstaltungen/2026-komso-academy#c722404 <p>The <a class="external-link" href="https://komso.org/activities/komso-academies/" target="_blank" title="KOMSO Academy">KOMSO Academy</a> aims to address novel business and technology trends at an early stage of development. It brings together leading experts from industry and academia to discuss the current state of the art, as well as potentials, risks, and future developments. The program is open to short contributions from participants seeking feedback on their specific problems.</p> <p>The present workshop focuses on deep learning concepts for solving partial differential equations and related parametric studies. We will host two external speakers presenting on advanced optimization schemes for faster training and on the use of deep learning in optimization processes. In addition, a concrete application for improving multiscale methods will be discussed, and reports on successful implementations in applied and industrial contexts will be presented.</p> <p>The program also includes hands-on training using a dedicated software toolbox (to be announced later). In the training sessions, participants will have the opportunity to gain direct experience with solving prepared example problems using deep learning approaches. PDE-based parameter studies and parameter identification problems will be considered.</p> Content content-722373 Mon, 04 May 2026 10:47:13 +0200 https://www.uni-bremen.de/techmath/veranstaltungen/2026-komso-academy#c722373 <p>Within the KOMSO Academy, we offer a hands-on workshop on deep learning approaches for partial differential equations (PDEs) and their practical implementation in computational frameworks. </p> <p>The course begins with an introduction to fundamental concepts for solving PDEs using deep learning techniques. Using classical benchmark problems such as Poisson and Darcy flow equations, participants will learn how physics-informed loss functions can incorporate physical laws directly into neural network training. Alongside these foundations, we emphasize hands-on implementation strategies within a flexible software environment that supports a wide variety of PDE-related problems, including those defined on complex and time-dependent domains.</p> <p>Building on these basics, the course then introduces operator learning methods for PDEs, which are particularly suited for parametric studies and parameter identification tasks. Participants will explore modern architectures such as PCA-based networks, DeepONets, and Fourier Neural Operators (FNOs), as well as their variants, gaining insight into both their theoretical foundations and their practical realization.</p> <p>In the final session, all participants come together to discuss their own PDE-related challenges. This interactive format allows for an exchange of ideas on suitable deep learning approaches and implementation strategies for specific applications. Participants are encouraged to bring their own problems for discussion.</p> Content content-722369 Mon, 11 May 2026 15:01:33 +0200 https://www.uni-bremen.de/techmath/veranstaltungen/2026-komso-academy#c722369 <p>The KOMSO Academy will take place at the <a href="https://www.bosch.de/unternehmen/bosch-in-deutschland/renningen/" title="Öffnet externen Link in neuem Fenster" target="_blank">Bosch Forschungscampus, Renningen/Stuttgart.</a></p> <p>For more detailed information about <strong>arriving by car or by public transportation</strong> from Stuttgart airport or Stuttgart main station, <a class="download-link" href="t3://file?uid=203638" target="_parent">click here</a>.</p> <p>Online participation is possible for a limited number of participants and the dial-in data will be provided by E-mail.</p> Content