CosmoStat_Twitter_Profile

CosmoSClub: 28-03-2018

Date: March 28th 2018

Speaker: Anais Moller (Australian National University)

Title: The next-generation of type Ia supernovae Dark Energy studies


Abstract

Currently Dark Energy studies with type Ia supernovae rely on a spectroscopically classified sample. The Dark Energy Survey (DES) is entering its last year of observations and a first cosmological analysis with the spectroscopically confirmed supernova sample is on the way. However, present and future surveys as DES and LSST will do cosmological analysis with a photometrically classified type Ia supernova sample. For this, a reliable photometric classification is necessary, which can process large number of candidates and obtain a high-purity sample. 
In this talk, I will first present preliminary cosmological parameter constraints from the first 3-years of the DES supernova survey. The sample is composed by  251 spectroscopically confirmed Type Ia Supernovae (0.02 < z < 0.9) discovered during the first 3 years of the Dark Energy Survey Supernova Program. I will also discuss about the future analysis with the DES 5-year photometric supernova sample. In particular, I will discuss a photometric classification method based on recurrent neural networks that can classify quickly large number of supernovae with high accuracy using only photometric measurements and time as input. This method includes a bayesian interpretation of classification probabilities which will be fundamental for a cosmology analysis. In addition, this method also classifies partial light-curves with high accuracy and speed which will allow to distribute resources towards promising candidates and can be applied to other transients.
CosmoStat_Twitter_Profile

CosmoSClub: 01-02-2018

Date: February 1st 2018

Speaker: Sandrine Codis (IAP)

Title: Intrinsic alignments : theoretical and numerical insights


Abstract

In this talk, I will show how both dark matter and hydrodynamical simulations predict that the morphology of galaxies is correlated with the cosmic web. This large-scale coherence of galaxy shapes could possibly induce some non-negligible level of contamination for future cosmic shear experiments. Because this effect is very sensitive to the small scale baryonic physics, it is difficult to use dark matter-only simulations as the sole resource to predict and control intrinsic alignments. I will show how state-of-the-art hydrodynamical simulations like Horizon-AGN can be used to shed light on the level of intrinsic alignment we should expect for future weak lensing measurements.  On the theoretical side, I will describe an analytical Lagrangian model that reproduces qualitatively the correlations between the intrinsic angular momentum of galaxies and the cosmic filaments. The key ingredient is to take into account the anisotropy of the cosmic web in the standard theory of spin acquisition by tidal torquing.

CosmoStat_Twitter_Profile

CosmoClub: 17-01-2018

Date: January 17th 2018

Speaker: François Orieux (L2S, Centrale-Supélec)

Title: Semi-unsupervised Bayesian Convex Image Restoration with Location Mixture of Gaussian


Abstract

Convex image restoration is a major field in inverse problems. The problem is often addressed by hand-tuning hyper-parameters. We propose an incremental contribution about a Bayesian approach where a convex field is constructed via Location Mixture of Gaussian and the estimator computed with a fast MCMC algorithm. Main contributions are a new field with several operator avoiding crosslike artifacts and a fallback sampling algorithm to prevent numerical errors. Results, in comparison to standard supervised results, have equivalent quality in a quasi-unsupervised approach and go with uncertainty quantification.

CosmoStat_Twitter_Profile

CosmoClub: 08-12-2017

Date: December 8th 2017

Speaker: Emilie Chouzenoux (Université Paris-Est)

Title: Proximity operator computation for video restoration [slides]


Abstract

Proximal methods have gained much interest for solving large-scale possibly non smooth optimization problems. When dealing with complicated convex functions, the expression of the proximity operator is however often non explicit and it thus needs to be determined numerically. We show in this work how block-coordinate algorithms can be designed to perform this task. We deduce also distributed optimization strategies allowing us to implement our solutions on multicore architectures. Applications of these methods to video restoration of old TV sequences illustrate the good performance of the proposed algorithms.