Speaker
Description
The BEST collaboration’s equation of state (EoS) maps a 3D Ising model onto the lattice QCD EoS but contains 4 free parameters related to the size, location, and spread of the critical region across the QCD phase diagram. However, certain combinations of those 4 free parameters lead to acausal ($c_s^2>1$) or unstable ($\chi_2^B<0$) realizations of the EoS that should not be considered. Here, we use an active learning framework to rule out pathological EoS efficiently. We show that checking stability and causality for a small fraction of the available parameter combinations is sufficient to produce algorithms that perform with >96% accuracy across the entire parameter space. Though we work with a specific case, this approach can be generalized to any model containing a parameter space-class correspondence.
