Timezone: »

 
Flow Matching for Scalable Simulation-Based Inference
Jonas Wildberger · Maximilian Dax · Simon Buchholz · Stephen R. Green · Jakob Macke · Bernhard Schölkopf

Sat Jul 29 02:30 PM -- 02:45 PM (PDT) @
in 2nd ICML Workshop on Machine Learning for Astrophysics »

Neural posterior estimation methods based on dis-crete normalizing flows have become establishedtools for simulation-based inference (SBI), butscaling them to high-dimensional problems can bechallenging. Building on recent advances in gen-erative modeling, we here present flow matchingposterior estimation (FMPE), a technique for SBIusing continuous normalizing flows. Like diffu-sion models, and in contrast to discrete flows, flowmatching allows for unconstrained architectures,providing enhanced flexibility for complex datamodalities. Flow matching, therefore, enablesexact density evaluation, fast training, and seam-less scalability to large architectures—making itideal for SBI. To showcase the improved scalabil-ity of our approach, we apply it to a challengingastrophysics problem: for gravitational-wave in-ference, FMPE outperforms methods based oncomparable discrete flows, reducing training timeby 30% with substantially improved accuracy

Author Information

Jonas Wildberger (Max Planck Institute for Intelligent Systems)
Maximilian Dax (Max Planck Institute for Intelligent Systems)
Simon Buchholz (Max Planck Institute for Intelligent Systems)
Stephen R. Green (University of Nottingham)
Jakob Macke (University of Tuebingen)
Bernhard Schölkopf (MPI for Intelligent Systems Tübingen, Germany)

Bernhard Scholkopf received degrees in mathematics (London) and physics (Tubingen), and a doctorate in computer science from the Technical University Berlin. He has researched at AT&T Bell Labs, at GMD FIRST, Berlin, at the Australian National University, Canberra, and at Microsoft Research Cambridge (UK). In 2001, he was appointed scientific member of the Max Planck Society and director at the MPI for Biological Cybernetics; in 2010 he founded the Max Planck Institute for Intelligent Systems. For further information, see www.kyb.tuebingen.mpg.de/~bs.

More from the Same Authors