LIGO Seminar

https://caltech.zoom.us/j/82072539167

Title: "Anomaly Detection with ML: from galaxy spectra to GW strain data" with Dovi Poznanski, Tel Aviv University

Abstract: Over the past few years my group has been focused on applying and developing machine learning tools that can help us find the things we did not know we should be looking for in large amounts of data. I will argue that this is a necessary step, with datasets ever growing in size and complexity. I will briefly discuss the lessons learned and the results obtained in an array of domains in astrophysics, from optical galaxy spectra to gravitational-waves strain data from LIGO. On the latter, I will show initial results from an ongoing project where we train a Deep Learning algorithm on the noise properties of the two LIGO detectors, in order to identify new sources as coincident "not-noise". This work may also promote our understanding of the noise properties of the detectors, in addition to opening an avenue for the discovery of sources with an unusual or unexpected morphology in time-frequency.