Abstract

A crucial step in a data release is performing data validation to distinguish true scientific anomalies from those caused by instrument or calibration errors. Gaia Data Release 4 (DR4) presents significant computational and methodological challenges due to its vast, high-dimensional datasets. One of the aspects of validation process is anomaly detection within the photometry data, which involves identifying deviations from expected patterns in time series data. In this talk, I will first address the broader data validation problem for epoch photometry in Gaia DR4 . I will then introduce a proposed anomaly detection pipeline, highlighting various techniques for detecting anomalies in time series data, including both statistical approaches and machine learning-based methods. By combining domain expertise with data-driven methodologies, this work aims to enhance the validation of Gaia’s photometric time series data.