Outlier Detection

We tried several approaches. The first thing that we tried was to use the depernin oteacordens. And it gave us fairly good results on the control data set. But then when we applied it for practical forilwol data coming from habe space telescopes, other telescopes, we didn't get the results that we wanted. Maybe because also the regular galaxies are very different from each other. There's a whole spectrum of galacy tat are considered regular, but they don't look alike at all. That might have relatively affected the performance of the auto encorders. In addition to some images that are out, the fact that something was wrong with the imagen.