Neural networks optimized by genetic algorithms in cosmology

The applications of artificial neural networks in the cosmological field have shone successfully during past decade, this is due to their great ability modeling large amounts datasets and complex nonlinear functions. However, some cases, use still remains controversial because ease producing inaccurate results when hyperparameters are not carefully selected. In paper, find optimal combination networks, we propose take advantage genetic algorithms. As a proof concept, analyze three different cases test performance architectures achieved with algorithms compare them standard process, consisting grid all possible configurations. First, carry out model-independent reconstruction distance modulus using type Ia supernovae compilation. Second, learn infer equation state for quintessence model, finally data from combined redshift catalog predict photometric given six bands (urgizy). We found that improve considerably generation network architectures, which can ensure more confidence physical better metrics respect method.