Philipp Brendel from Fraunhofer IISB attended the fifth edition of the Mathematical and Scientific Machine Learning Conference 2025 (MSML), held from August 4–8. He presented a poster on the FASTEST project at the event, which focuses on advancing the mathematical theory and algorithms of machine learning, as well as their applications in scientific computing and engineering. This year’s conference covered a wide range of themes, including attention mechanisms, causal inference, optimization, optimal transport, and graph neural networks. Additional topics featured included neural operators, transformers, state-space models, and the broader mathematics underlying machine learning.