DEEP Learning For Large Deep Imaging Programs – DEEPDIP
Prepare the dataset, develop the deep learning methods and apply them to scientific questions.
Tools for data production, archiving and access are operational
Continuation of the work
1. “The NewHorizon Simulation - To Bar Or Not To Bar” Reddish, Kraljic et al., submitted to MNRAS 2021
2. “Cosmic filaments delay quenching inside clusters” Kotecha et al., submitted to MNRAS 2021
3. “Gas accretion and Ram Pressure Stripping of Halos in Void Walls” Thompson, Smith, Kraljic, submitted to MNRAS 2021
4. “The role of the cosmic web in the scatter of the galaxy stellar mass - gas metallicity relation” Donnan, Tojeiro, Kraljic, submit-ted to Nature Astronomy 2021
5. “BCG alignment with the Locations of Cluster Satellites and the Large Scale Structure out to 10 R200,Paper I: General Depend-encies and Systematics” Smith, Hwang, Kraljic et al., submitted to ApJ 2021
6. « PhotoWeb redshift: boosting photometric redshift accuracy with large spectroscopic surveys » Shuntov, Pasquet, Arnouts et al., A&A 2020
In future large photometric surveys like LSST or Euclid, high quality photometric redshift measurements and light curve classification will play a central role. We propose to revisit the current methodology, which is based on extracting a small set of photometric features used as input for the SED fitting or standard machine learning techniques.By taking advantage of the latest Deep Learning techniques, the GPU acceleration and the ever growing size of spectroscopic samples, we can bypass the current limitations and deal directly with multi-band images at the pixel level (i.e. without photometric feature extractions). Our ambition is to improve, through these new methods, the accuracy of cosmological analyses using the Hubble diagram of supernovae and the measurement of the evolution of large scale structures. In addition, we will provide the community with the most accurate photometric redshifts as well as tools to ensure the optimal exploitation of the next generation of surveys.
Project coordination
Dominique Fouchez (centre national de la recherche scientifiqueDelegation provence et Corse_Centre de physique des particules de Marseille)
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
Partner
IAP Institut d'astrophysique de Paris
UM-LIRMM Laboratoire d'Informatique, de Robotique et de Microélectronique de Montpellier
LAM Laboratoire d'astrophysique de Marseille
CNRS DR12_CPPM centre national de la recherche scientifiqueDelegation provence et Corse_Centre de physique des particules de Marseille
Help of the ANR 632,682 euros
Beginning and duration of the scientific project:
November 2019
- 48 Months
