The Asia tour of our Founder continues

In July, Lorenz Richter presented at DO x ML 2026, the 8th Conference on Discrete Optimization and Machine Learning, in Tokyo. He then traveled to Taipei for ICOML 2026, the International Conference on Optimization and Machine Learning.

Both talks explored "The duality between sampling and optimization: New perspectives from diffusions and optimal control"  – how ideas from stochastic optimal control and diffusion-based methods can be used to tackle high-dimensional sampling and non-convex optimization problems. The theme builds on research Lorenz developed together with Julius Berner and Karen Ullrich, connecting stochastic optimal control to diffusion-based generative modeling, work that showed, among other results, that the evidence lower bound used to train diffusion models follows directly from the verification theorem in control theory.

It's great to see dida represented in these research communities through Lorenz's work. Contributing to fundamental ML research is something we care about and are happy to be part of.