This document summarizes a presentation on analyzing astronomical time series data using information theoretic learning approaches. It discusses challenges with irregularly sampled and noisy light curve data from astronomical surveys. It proposes using correntropy, a generalized correlation measure, within a periodic kernel to create a Correntropy Kernelized Periodogram (CKP) for discriminating periodic vs non-periodic light curves and estimating periods of periodic curves. It applies this approach to real survey data from MACHO and EROS, achieving high classification accuracy and ability to process billions of light curves efficiently using GPU clusters.