Automating Inference of Binary Microlensing Events with Neural Density Estimation. (arXiv:2010.04156v1 [astro-ph.IM])
<a href="http://arxiv.org/find/astro-ph/1/au:+Zhang_K/0/1/0/all/0/1">Keming Zhang</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Bloom_J/0/1/0/all/0/1">Joshua S. Bloom</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Gaudi_B/0/1/0/all/0/1">B. Scott Gaudi</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Lanusse_F/0/1/0/all/0/1">Francois Lanusse</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Lam_C/0/1/0/all/0/1">Casey Lam</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Lu_J/0/1/0/all/0/1">Jessica Lu</a>
Automated inference of binary microlensing events with traditional
sampling-based algorithms such as MCMC has been hampered by the slowness of the
physical forward model and the pathological likelihood surface. Current
analysis of such events requires both expert knowledge and large-scale grid
searches to locate the approximate solution as a prerequisite to MCMC posterior
sampling. As the next generation, space-based microlensing survey with the
Roman Space Observatory is expected to yield thousands of binary microlensing
events, a new scalable and automated approach is desired. Here, we present an
automated inference method based on neural density estimation (NDE). We show
that the NDE trained on simulated Roman data not only produces fast, accurate,
and precise posteriors but also captures expected posterior degeneracies. A
hybrid NDE-MCMC framework can further be applied to produce the exact
posterior.
Automated inference of binary microlensing events with traditional
sampling-based algorithms such as MCMC has been hampered by the slowness of the
physical forward model and the pathological likelihood surface. Current
analysis of such events requires both expert knowledge and large-scale grid
searches to locate the approximate solution as a prerequisite to MCMC posterior
sampling. As the next generation, space-based microlensing survey with the
Roman Space Observatory is expected to yield thousands of binary microlensing
events, a new scalable and automated approach is desired. Here, we present an
automated inference method based on neural density estimation (NDE). We show
that the NDE trained on simulated Roman data not only produces fast, accurate,
and precise posteriors but also captures expected posterior degeneracies. A
hybrid NDE-MCMC framework can further be applied to produce the exact
posterior.
http://arxiv.org/icons/sfx.gif
