Abstract:
|
The Large Synoptic Survey Telescope (LSST) will place great demands on data analysis procedures, from a combination of the quantity and complexity of the data that will result. These data are still a few years away, so ongoing work to develop analysis pipelines requires realistic simulation software. One particular problem is the simulation of galaxy images in a manner that appropriately samples the space of shapes and contamination effects, at the resolution and error level of LSST, and possibly conditional on other properties of the galaxy. Here we will present an approach to this problem that builds on recent work in the use of Wasserstein distance as a means of interpolation between training images. This work is joint with Benjamin LeRoy.
|