Seismic Inference using Genetic Algorithms

Abstract

A flood of reliable seismic data will soon arrive. The migration to largertelescopes on the ground may free up 4-m class instruments for multi-sitecampaigns, and several forthcoming satellite missions promise to yieldnearly uninterrupted long-term coverage of many pulsating stars. We willthen face the challenge of determining the fundamental properties of thesestars from the data, by trying to match them with the output of ourcomputer models. The traditional approach to this task is to make informedguesses for each of the model parameters, and then adjust them iterativelyuntil an adequate match is found. The trouble is: how do we know that oursolution is unique, or that some other combination of parameters will notdo even better? Computers are now sufficiently powerful and inexpensivethat we can produce large grids of models and simply compare all ofthem to the observations. The question then becomes: what range ofparameters do we want to consider, and how many models do we want tocalculate? This can minimize the subjective nature of the process, but itmay not be the most efficient approach and it may give us a false sense ofsecurity that the final result is correct, when it is really justoptimal. I discuss these issues in the context of recent advances inthe asteroseismological analysis of white dwarf stars.