Scalable backpropagation for Gaussian Processes using celerite

Preprint English OPEN
Foreman-Mackey, Daniel;
(2018)
  • Subject: Astrophysics - Instrumentation and Methods for Astrophysics

This research note presents a derivation and implementation of efficient and scalable gradient computations using the celerite algorithm for Gaussian Process (GP) modeling. The algorithms are derived in a "reverse accumulation" or "backpropagation" framework and they ca... View more
  • References (3)

    Abadi, M., et al. 2016, in OSDI, Vol. 16, 265-283 Ambikasaran, S. 2015, Numer. Linear Algebra Appl., 22, 1102

    Angus, R., et al. 2018, MNRAS, 474, 2094 Bond, J. R., & Efstathiou, G. 1987, MNRAS, 226, 655 Brewer, B. J., & Stello, D. 2009, MNRAS, 395, 2226 Carpenter, B., et al. 2015, ArXiv, 1509.07164 Czekala, I., et al. 2017, ApJ, 840, 49 Foreman-Mackey, D. 2018, celerite-grad, , , doi:10.5281/zenodo.1160614 Giles, M. 2008, An extended collection of matrix derivative results for forward and reverse mode automatic differentiation, Tech. Rep. NA-08/01, Oxford University Computing Laboratory

    Haywood, R. D., et al. 2014, MNRAS, 443, 2517 Kelly, B. C., et al. 2014, ApJ, 788, 33 LeCun, Y., et al. 2015, Nature, 521, 436 Luger, R., et al. 2016, AJ, 152, 100 Murray, I. 2016, ArXiv, 1602.07527 Rajpaul, V., et al. 2015, MNRAS, 452, 2269 Rasmussen, C. E., & Williams, K. I. 2006, Gaussian Processes for Machine Learning (MIT Press) Wandelt, B. D., & Hansen, F. K. 2003, PhRvD, 67, 023001 Wang, Y., et al. 2012, ApJ, 756, 67

  • Related Research Results (1)
    Inferred by OpenAIRE
    software
    Dfm/Celerite-Grad: Celerite-Grad V0.0.1 (2018)
    73%
  • Metrics
Share - Bookmark