skip to main content
DOE PAGES title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Uncertainties in Parameters Estimated with Neural Networks: Application to Strong Gravitational Lensing

Abstract

In Hezaveh et al. (2017) we showed that deep learning can be used for model parameter estimation and trained convolutional neural networks to determine the parameters of strong gravitational lensing systems. Here we demonstrate a method for obtaining the uncertainties of these parameters. We review the framework of variational inference to obtain approximate posteriors of Bayesian neural networks and apply it to a network trained to estimate the parameters of the Singular Isothermal Ellipsoid plus external shear and total flux magnification. We show that the method can capture the uncertainties due to different levels of noise in the input data, as well as training and architecture-related errors made by the network. To evaluate the accuracy of the resulting uncertainties, we calculate the coverage probabilities of marginalized distributions for each lensing parameter. By tuning a single hyperparameter, the dropout rate, we obtain coverage probabilities approximately equal to the confidence levels for which they were calculated, resulting in accurate and precise uncertainty estimates. Our results suggest that neural networks can be a fast alternative to Monte Carlo Markov Chains for parameter uncertainty estimation in many practical applications, allowing more than seven orders of magnitude improvement in speed.

Authors:
ORCiD logo [1]; ORCiD logo [1]; ORCiD logo [1]
  1. Stanford Univ. (United States). Kavli Inst. for Particle Astrophysics and Cosmology; SLAC National Accelerator Lab., Menlo Park, CA (United States)
Publication Date:
Research Org.:
SLAC National Accelerator Lab., Menlo Park, CA (United States)
Sponsoring Org.:
USDOE
OSTI Identifier:
1419653
Grant/Contract Number:  
AC02-76SF00515
Resource Type:
Accepted Manuscript
Journal Name:
The Astrophysical Journal. Letters
Additional Journal Information:
Journal Volume: 850; Journal Issue: 1; Journal ID: ISSN 2041-8213
Publisher:
Institute of Physics (IOP)
Country of Publication:
United States
Language:
English

Citation Formats

Perreault Levasseur, Laurence, Hezaveh, Yashar D., and Wechsler, Risa H.. Uncertainties in Parameters Estimated with Neural Networks: Application to Strong Gravitational Lensing. United States: N. p., 2017. Web. https://doi.org/10.3847/2041-8213/aa9704.
Perreault Levasseur, Laurence, Hezaveh, Yashar D., & Wechsler, Risa H.. Uncertainties in Parameters Estimated with Neural Networks: Application to Strong Gravitational Lensing. United States. https://doi.org/10.3847/2041-8213/aa9704
Perreault Levasseur, Laurence, Hezaveh, Yashar D., and Wechsler, Risa H.. Wed . "Uncertainties in Parameters Estimated with Neural Networks: Application to Strong Gravitational Lensing". United States. https://doi.org/10.3847/2041-8213/aa9704. https://www.osti.gov/servlets/purl/1419653.
@article{osti_1419653,
title = {Uncertainties in Parameters Estimated with Neural Networks: Application to Strong Gravitational Lensing},
author = {Perreault Levasseur, Laurence and Hezaveh, Yashar D. and Wechsler, Risa H.},
abstractNote = {In Hezaveh et al. (2017) we showed that deep learning can be used for model parameter estimation and trained convolutional neural networks to determine the parameters of strong gravitational lensing systems. Here we demonstrate a method for obtaining the uncertainties of these parameters. We review the framework of variational inference to obtain approximate posteriors of Bayesian neural networks and apply it to a network trained to estimate the parameters of the Singular Isothermal Ellipsoid plus external shear and total flux magnification. We show that the method can capture the uncertainties due to different levels of noise in the input data, as well as training and architecture-related errors made by the network. To evaluate the accuracy of the resulting uncertainties, we calculate the coverage probabilities of marginalized distributions for each lensing parameter. By tuning a single hyperparameter, the dropout rate, we obtain coverage probabilities approximately equal to the confidence levels for which they were calculated, resulting in accurate and precise uncertainty estimates. Our results suggest that neural networks can be a fast alternative to Monte Carlo Markov Chains for parameter uncertainty estimation in many practical applications, allowing more than seven orders of magnitude improvement in speed.},
doi = {10.3847/2041-8213/aa9704},
journal = {The Astrophysical Journal. Letters},
number = 1,
volume = 850,
place = {United States},
year = {2017},
month = {11}
}

Journal Article:
Free Publicly Available Full Text
Publisher's Version of Record

Citation Metrics:
Cited by: 3 works
Citation information provided by
Web of Science

Save / Share:

Works referenced in this record:

Dermatologist-level classification of skin cancer with deep neural networks
journal, January 2017

  • Esteva, Andre; Kuprel, Brett; Novoa, Roberto A.
  • Nature, Vol. 542, Issue 7639
  • DOI: 10.1038/nature21056

Fast automated analysis of strong gravitational lenses with convolutional neural networks
journal, August 2017

  • Hezaveh, Yashar D.; Levasseur, Laurence Perreault; Marshall, Philip J.
  • Nature, Vol. 548, Issue 7669
  • DOI: 10.1038/nature23463

Finding strong lenses in CFHTLS using convolutional neural networks
journal, June 2017

  • Jacobs, C.; Glazebrook, K.; Collett, T.
  • Monthly Notices of the Royal Astronomical Society, Vol. 471, Issue 1
  • DOI: 10.1093/mnras/stx1492

Human-level control through deep reinforcement learning
journal, February 2015

  • Mnih, Volodymyr; Kavukcuoglu, Koray; Silver, David
  • Nature, Vol. 518, Issue 7540
  • DOI: 10.1038/nature14236

Finding strong gravitational lenses in the Kilo Degree Survey with Convolutional Neural Networks
journal, August 2017

  • Petrillo, C. E.; Tortora, C.; Chatterjee, S.
  • Monthly Notices of the Royal Astronomical Society, Vol. 472, Issue 1
  • DOI: 10.1093/mnras/stx2052

Mastering the game of Go with deep neural networks and tree search
journal, January 2016

  • Silver, David; Huang, Aja; Maddison, Chris J.
  • Nature, Vol. 529, Issue 7587
  • DOI: 10.1038/nature16961

The sl2s Galaxy-Scale lens Sample. v. dark Matter Halos and Stellar imf of Massive Early-Type Galaxies out to Redshift 0.8
journal, February 2015

  • Sonnenfeld, Alessandro; Treu, Tommaso; Marshall, Philip J.
  • The Astrophysical Journal, Vol. 800, Issue 2
  • DOI: 10.1088/0004-637X/800/2/94

    Works referencing / citing this record:

    Surveying the reach and maturity of machine learning and artificial intelligence in astronomy
    journal, September 2019

    • Fluke, Christopher J.; Jacobs, Colin
    • WIREs Data Mining and Knowledge Discovery, Vol. 10, Issue 2
    • DOI: 10.1002/widm.1349

    Gaia GraL: Gaia DR2 gravitational lens systems : II. The known multiply imaged quasars⋆
    journal, October 2018


    Is every strong lens model unhappy in its own way? Uniform modelling of a sample of 13 quadruply+ imaged quasars
    journal, December 2018

    • Shajib, A. J.; Birrer, S.; Treu, T.
    • Monthly Notices of the Royal Astronomical Society, Vol. 483, Issue 4
    • DOI: 10.1093/mnras/sty3397

    Direct detection of dark matter substructure in strong lens images with convolutional neural networks
    journal, January 2020


    Deep neural networks to enable real-time multimessenger astrophysics
    journal, February 2018


    Cosmological constraints from noisy convergence maps through deep learning
    journal, December 2018


    An Ensemble of Bayesian Neural Networks for Exoplanetary Atmospheric Retrieval
    journal, June 2019

    • Cobb, Adam D.; Himes, Michael D.; Soboczenski, Frank
    • The Astronomical Journal, Vol. 158, Issue 1
    • DOI: 10.3847/1538-3881/ab2390

    Nonparametric Star Formation History Reconstruction with Gaussian Processes. I. Counting Major Episodes of Star Formation
    journal, July 2019

    • Iyer, Kartheik G.; Gawiser, Eric; Faber, Sandra M.
    • The Astrophysical Journal, Vol. 879, Issue 2
    • DOI: 10.3847/1538-4357/ab2052

    Data-driven Reconstruction of Gravitationally Lensed Galaxies Using Recurrent Inference Machines
    journal, September 2019

    • Morningstar, Warren R.; Levasseur, Laurence Perreault; Hezaveh, Yashar D.
    • The Astrophysical Journal, Vol. 883, Issue 1
    • DOI: 10.3847/1538-4357/ab35d7

    Mining for Dark Matter Substructure: Inferring Subhalo Population Properties from Strong Lenses with Machine Learning
    journal, November 2019

    • Brehmer, Johann; Mishra-Sharma, Siddharth; Hermans, Joeri
    • The Astrophysical Journal, Vol. 886, Issue 1
    • DOI: 10.3847/1538-4357/ab4c41

    A Machine Learning Based Morphological Classification of 14,245 Radio AGNs Selected from the Best–Heckman Sample
    journal, February 2019

    • Ma, Zhixian; Xu, Haiguang; Zhu, Jie
    • The Astrophysical Journal Supplement Series, Vol. 240, Issue 2
    • DOI: 10.3847/1538-4365/aaf9a2