Save 37% off PRO during our Black Friday Sale! »

Image/data processing for direct imaging of exoplanets (PyAstro2018)

Image/data processing for direct imaging of exoplanets (PyAstro2018)

Lightning talk presented during the Python in Astronomy 2018 workshop @ Center for Computational Astrophysics, Flatiron Institute, New York. The slides present two Python packages for direct imaging of exoplanets: VIP and SODINN. VIP is an package for differential imaging and high-contrast imaging data analysis (https://github.com/vortex-exoplanet/VIP). SODINN is focused on deep learning applied to pattern recognition for detection of exoplanets in high-contrast imaging sequences.

Transcript

  1. Carlos Alberto Gomez Gonzalez Python in Astronomy, 01/05/2018 Vortex Image

    Processing Image/data processing for direct imaging of exoplanets SODINN: supervised direct detection of exoplanets
  2. • https://github.com/vortex-exoplanet/VIP • Available on Pypi: >>> pip install vip_hci

    • Python 2/3 compatibility • Documentation (Sphinx): http://vip.readthedocs.io/ • 20k lines of code, 10 contributors, 15+ citations, ~40 users Gomez Gonzalez et al. 2017 Vortex Image Processing library
  3. Vortex Image Processing library Gomez Gonzalez et al. 2017 •

    Travis & Pytest (improving coverage)
  4. planet Angular differential imaging - bright synthetic planet 12 imaged

    exoplanets out of 3725 confirmed candidates!!! Image sequence animation
  5. Marois et al. 2007 Gomez Gonzalez et al. 2016 Lafrenière

    et al. 2007 Marois et al. 2007 Soummer et al. 2012 Amara & Quanz 2012 Absil et al. 2013 Gomez Gonzalez et al. 2017 Marois et al. 2014 Marois et al. 2014 Hagelberg et al. 2015 Mugnier et al. 2009 Cantalloube et al. 2015 ZOO of post-processing algorithms
  6. • Jupyter tutorial(s) • Workflows & reproducibility https://github.com/carlgogo/vip-tutorial Tutorial

  7. API

  8. API

  9. API

  10. Unsupervised Supervised PC 1 PC 2 Dimensionality reduction Clustering Machine

    learning in a nutshell and reinforcement learning Density estimation Regression Classification
  11. Training set (labeled data) { Network architecture Loss function and

    regularization { Optimization
  12. Input X 1st Layer (data transformation) 2nd Layer (data transformation)

    Nth Layer (data transformation) … Predictions Y’ Labels Y Loss function weights weights weights Optimizer loss score weight update
  13. Supervised detection of exoplanets Gomez Gonzalez et al. 2018

  14. Quite often, there's no need to re-invent the wheel

  15. ¡Gracias! carlos.gomez@univ-grenoble-alpes.fr carlgogo carlosalbertogomezgonzalez https://carlgogo.github.io/