Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 96: Data, AI, Computing 7 (uncertainties, likelihoods)

T 96.7: Vortrag

Donnerstag, 7. März 2024, 17:30–17:45, Geb. 30.33: MTI

dilax: Differentiable Binned Likelihoods in JAX — Peter Fackeldey, Benjamin Fischer, •Felix Zinn, and Martin Erdmann — III. Physikalisches Institut A, RWTH Aachen University

dilax is a software package for statistical inference using likelihood functions of binned data. It fulfils three key concepts: performance, differentiability, and object-oriented statistical model building.

dilax is build on JAX - a powerful autodifferentiation Python framework. By making every component in dilax a “PyTree”, each component can be jit-compiled (jax.jit), vectorized (jax.vmap) and differentiated (jax.grad). This enables additionally novel computational concepts, such as running thousands of fits simultaneously on a GPU.

We present the key concepts of dilax, show its features, and discuss performance benchmarks with toy datasets.

Keywords: Binned Likelihood; Fitting; JAX; Autodifferentiation; GPU