Calibrating cosmological simulations with implicit likelihood inference using galaxy growth observables by Yongseok Jo et al. on Wednesday 30 NovemberIn a novel approach employing implicit likelihood inference (ILI), also knownas likelihood-free inference, we calibrate the parameters of cosmologicalhydrodynamic simulations against observations, which has previously beenunfeasible due to the high computational cost of these simulations. Forcomputational efficiency, we train neural networks as emulators on ~1000cosmological simulations from the CAMELS project to estimate simulatedobservables, taking as input the cosmological and astrophysical parameters, anduse these emulators as surrogates to the cosmological simulations. Using thecosmic star formation rate density (SFRD) and, separately, stellar massfunctions (SMFs) at different redshifts, we perform ILI on selectedcosmological and astrophysical parameters (Omega_m, sigma_8, stellar windfeedback, and kinetic black hole feedback) and obtain full 6-dimensionalposterior distributions. In the performance test, the ILI from the emulatedSFRD (SMFs) can recover the target observables with a relative error of 0.17%(0.4%). We find that degeneracies exist between the parameters inferred fromthe emulated SFRD, confirmed with new full cosmological simulations. We alsofind that the SMFs can break the degeneracy in the SFRD, which indicates thatthe SMFs provide complementary constraints for the parameters. Further, we findthat the parameter combination inferred from an observationally-inferred SFRDreproduces the target observed SFRD very well, whereas, in the case of theSMFs, the inferred and observed SMFs show significant discrepancies thatindicate potential limitations of the current galaxy formation modeling andcalibration framework, and/or systematic differences and inconsistenciesbetween observations of the stellar mass function.arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.16461v1