Seminars

 

Giovedi 24 Gennaio 2019, h. 11:30 
 Aula L. Rosino Dipartimento di Fisica e Astronomia

 

  Julian Merten 

   INAF OAS, Bologna
 
 

 Cosmological applications of machine learning

 

 

 ABSTRACT

With the ever-increasing data volumes associated with upcoming astronomical surveys and numerical simulations, machine learning has become an attractive framework to not only handle, but optimally exploit all the information in such data. Especially deep learning techniques provide a unique way of describing data in a particularly flexible --data-driven-- way by constructing complex models from a deep chain of simple mathematical operations. In my presentation, I will provide a concise introduction to machine and deep learning and go through a number of applications in the field of cosmology. We will use large computer vision data vectors, convolutional neural networks and a number of different classifiers, including fully-connected neural network, to characterise weak lensing mass maps. In the time-domain we will apply long short-term memory networks to filter radio frequency interference from 21cm radio data and close with a look into the exciting field of generative models as a particular cheap way to produce Mock data. Different forms of generative adversarial networks are deployed in order to create an arbitrary number of galaxy images and galaxy cluster realisations by learning their principal representation from given training sets.

 

  

 

          Previous Seminars                                             Next seminars  

__________________________________________________________________________________

       To propose a seminar                                              To be included send e-mail to:seminar

We use cookies to improve our website and your experience when using it. Cookies used for the essential operation of the site have already been set. To find out more about the cookies we use and how to delete them, see our privacy policy.

I accept cookies from this site.

EU Cookie Directive Module Information