GAN for Time Series Data Augmentation in Astronomy

Real-world datasets are often imbalanced, which is a problem for training CNNs. Data augmentation techniques have been proposed for image recognition tasks, but only a few have been developed for time series. I will describe a conditional Wasserstein GAN. Our model can learn the implicit probability distribution of a dataset conditioned on the irregular sampling times, amplitudes, and class of the time series, and generate a variety of realistic samples to complement the original dataset.
Cover Image

Pavlos is the Scientific Program Director,Institute for Applied Computational Science (IACS) at Harvard University, andleads the Data Science Masters Program at Harvard. Pavlos has had adistinguished career as a scientist and data science educator, and todayteaches the CS109 series for basic and advanced data science at Harvard, aswell as the capstone course (industry-sponsored data science projects) for theIACS Masters Program. Pavlos has a Ph.D in theoretical physics from theUniversity of Pennsylvania. Pavlos research has since branched into the use ofmachine learning and AI in astronomy, and computer science. He is excited to bea partner of, helping steer us to world-class excellence in AI Researchand education, this summer and beyond, sets sail on its mission.

Share this post