On the deduction of galaxy abundances with evolutionary neural networks

A growing number of indicators are now being used with someconfidence to measure the metallicity(Z) of photoionisation regions in planetarynebulae, galactic HII regions(GHIIRs), extra-galactic HII regions(EGHIIRs) andHII galaxies(HIIGs). However, a universal indicator valid also at high metallic-ities has yet to be found. Here, we report on a new artificial intelligence-basedapproach to determine metallicity indicators that shows promise for the provisionof improved empirical fits. The method hinges on the application of an evolu-tionary neural network to observational emission line data. The network’s DNA,encoded in its architecture, weights and neuron transfer functions, is evolved us-ing a genetic algorithm. Furthermore, selection, operating on a set of 10 distinctneuron transfer functions, means that the empirical relation encoded in the net-work solution architecture is in functional rather than numerical form. Thus thenetwork solutions providean equationfor the metallicity in terms of line ratioswithoutaprioriassumptions. Tapping into the mathematical power offeredby this approach, we applied the network to detailed observations of both neb-ula and auroral emission lines from 0.33μm−1μmfor a sample of 96 HII-typeregions and we were able to obtain an empirical relation betweenZandS23with a dispersion of only 0.16 dex. We show how the method can be used toidentify new diagnostics as well as the nonlinear relationship supposed to existbetween the metallicityZ, ionisation parameterUand effective (or equivalent)temperatureT∗