22–26 Jun 2026
Richard Roberts Auditorium
Europe/London timezone

Interpretable Machine Learning for Dynamical Dark Energy: From Phenomenology to String-Inspired Models

26 Jun 2026, 10:10
20m
Lecture Theatre B (Hicks Building)

Lecture Theatre B

Hicks Building

Speaker

Ivonne Zavala

Description

I will introduce SPIDER (Symbolic regression PIpeline for Dark Energy Reconstruction), an interpretable machine learning pipeline interfacing Exhaustive Symbolic Regression with a full Boltzmann solver to reconstruct dark energy properties directly from data. Applied to VCDM - a minimally modified gravity model - SPIDER uncovers remarkably simple analytic expressions for w(a), including a striking square-root parametrisation that improves on standard CPL. I will also present a string-motivated framework of time-dependent quintessence potentials as an effective description for coupled dark energy, where phantom crossing arises effectively, and discuss its reconstruction with SPIDER.

Presentation materials

There are no materials yet.