Université de GenèveDépartement de Physique ThéoriqueCAP Genève

Mining cosmic datasets

Date: 
17. January 2020 - 11:30
Speaker: 
Alireza Vafaei Sadr (IPM, Teheran)
New generations of observations and simulations provide a huge amount of data, posing the big data problem, which makes scientists consider data science as an increasingly important player in almost all data-based projects. There are a lot of interesting ideas about applications of data science tools in astrophysics and cosmology. As a particular application, anomaly/outlier detection is one challenging area. An example of application is provided in astronomy by the LSST and SKA, the next-generation optical and radio telescopes which are expected to observe completely new types of celestial objects lurking in the torrent of data in the 100PB-10EB range. In this presentation, we will know why anomaly detection is important. Then I will introduce a general anomaly detection framework based on dimensionality reduction and unsupervised clustering (DRAMA). This approach identifies the primary prototypes in the data with anomalies detected by their large distances from the prototypes, either in the latent space or in the original, high-dimensional space. Finally, we will see examples of how it works! The talk is based on https://arxiv.org/abs/1909.04060
Additional material: 

Address

Département de Physique Théorique
Université de Genève
24, quai Ernest Ansermet
1211 Genève 4
Switzerland
Directions & contact