Ponente
Descripción
Cosmological tensions between early- and late-Universe probes, most notably the Hubble tension in $H_0$ have intensified scrutiny of the ΛCDM paradigm and motivated flexible dark-energy scenarios beyond a cosmological constant. A persistent obstacle is that the functional form of the dark-energy equation of state $w(a)$ is usually chosen by hand (e.g.wCDM or CPL), so potentially viable models may never be tested.
We present an automated pipeline that searches a large combinatorial space of dynamical dark-energy parametrisations using a mask-based genetic algorithm coupled to the CLASS Boltzmann code. Each candidate is a sum of coefficient–shape terms $w(a)=\sum c_i f_j(a)$, the genetic algorithm selects which terms are active, while coefficients are constrained with datasets via Cobaya (with optional neural-network acceleration through CLiENT). Fitness is driven by $\chi^2$, enabling evolutionary discovery of competitive $w(a)$ structures without fixing a single parametric ansatz a priori.
This approach sits at the intersection of data-analysis and fundamental-physics themes: it uses evolutionary optimisation and likelihood-driven model search to explore dark-energy phenomenology that may help interpret, and ultimately alleviate, tensions with the concordance model. We discuss validation against standard baselines (including CPL), the best-performing symbolic forms found so far, and the path toward multi-probe analyses (CMB, BAO/DESI) needed to assess tension relief robustly.
