Up to two billion times acceleration of scientific simulations with deep neural architecture search

CoRR abs/2001.08055 (2020)

Authors:

MF Kasim, D Watson-Parris, L Deaconu, S Oliver, P Hatfield, DH Froula, G Gregori, M Jarvis, S Khatiwala, J Korenaga, J Topp-Mugglestone, E Viezzer, SM Vinko

Abstract:

Computer simulations are invaluable tools for scientific discovery. However, accurate simulations are often slow to execute, which limits their applicability to extensive parameter exploration, large-scale data analysis, and uncertainty quantification. A promising route to accelerate simulations by building fast emulators with machine learning requires large training datasets, which can be prohibitively expensive to obtain with slow simulations. Here we present a method based on neural architecture search to build accurate emulators even with a limited number of training data. The method successfully accelerates simulations by up to 2 billion times in 10 scientific cases including astrophysics, climate science, biogeochemistry, high energy density physics, fusion energy, and seismology, using the same super-architecture, algorithm, and hyperparameters. Our approach also inherently provides emulator uncertainty estimation, adding further confidence in their use. We anticipate this work will accelerate research involving expensive simulations, allow more extensive parameters exploration, and enable new, previously unfeasible computational discovery.

Full orbital solution for the binary system in the northern Galactic disc microlensing event Gaia16aye⋆

Astronomy & Astrophysics EDP Sciences 633 (2020) a98

Authors:

Ł Wyrzykowski, P Mróz, KA Rybicki, M Gromadzki, Z Kołaczkowski, M Zieliński, P Zieliński, N Britavskiy, A Gomboc, K Sokolovsky, ST Hodgkin, L Abe, GF Aldi, A AlMannaei, G Altavilla, A Al Qasim, GC Anupama, S Awiphan, E Bachelet, V Bakış, S Baker, S Bartlett, P Bendjoya, K Benson, IF Bikmaev, G Birenbaum, N Blagorodnova, S Blanco-Cuaresma, S Boeva, AZ Bonanos, V Bozza, DM Bramich, I Bruni, RA Burenin, U Burgaz, T Butterley, HE Caines, DB Caton, S Calchi Novati, JM Carrasco, A Cassan, V Čepas, M Cropper, M Chruślińska, G Clementini, A Clerici, D Conti, M Conti, S Cross, F Cusano, G Damljanovic, A Dapergolas, G D’Ago, JHJ de Bruijne, M Dennefeld, VS Dhillon, M Dominik, J Dziedzic, O Erece, MV Eselevich, H Esenoglu, L Eyer, R Figuera Jaimes, SJ Fossey, AI Galeev, SA Grebenev, AC Gupta, AG Gutaev, N Hallakoun, A Hamanowicz, C Han, B Handzlik, JB Haislip, L Hanlon, LK Hardy, DL Harrison, HJ van Heerden, VL Hoette, K Horne, R Hudec, M Hundertmark, N Ihanec, EN Irtuganov, R Itoh, P Iwanek, MD Jovanovic, R Janulis, M Jelínek, E Jensen, Z Kaczmarek, D Katz, IM Khamitov, Y Kilic, J Klencki, U Kolb, G Kopacki, VV Kouprianov, K Kruszyńska, S Kurowski, G Latev, C-H Lee, S Leonini, G Leto, F Lewis, Z Li, A Liakos, SP Littlefair, J Lu, CJ Manser, S Mao, D Maoz, A Martin-Carrillo, JP Marais, M Maskoliūnas, JR Maund, PJ Meintjes, SS Melnikov, K Ment, P Mikołajczyk, M Morrell, N Mowlavi, D Moździerski, D Murphy, S Nazarov, H Netzel, R Nesci, C-C Ngeow, AJ Norton, EO Ofek, E Pakštienė, L Palaversa, A Pandey, E Paraskeva, M Pawlak, MT Penny, BE Penprase, A Piascik, JL Prieto, JKT Qvam, C Ranc, A Rebassa-Mansergas, DE Reichart, P Reig, L Rhodes, J-P Rivet, G Rixon, D Roberts, P Rosi, DM Russell, R Zanmar Sanchez, G Scarpetta, G Seabroke, BJ Shappee, R Schmidt, Y Shvartzvald, M Sitek, J Skowron, M Śniegowska, C Snodgrass, PS Soares, B van Soelen, ZT Spetsieri, A Stankevičiūtė, IA Steele, RA Street, J Strobl, E Strubble, H Szegedi, LM Tinjaca Ramirez, L Tomasella, Y Tsapras, D Vernet, S Villanueva, O Vince, J Wambsganss, IP van der Westhuizen, K Wiersema, D Wium, RW Wilson, A Yoldas, R Ya Zhuchkov, DG Zhukov, J Zdanavičius, S Zoła, A Zubareva

The variable radio counterpart of Swift J1858.6-0814

Monthly Notices of the Royal Astronomical Society 496:4 (2020) 4127-4140

Authors:

van den Eijnden, J.; Degenaar, N.; Russell, T. D.; Buisson, D. J. K.; Altamirano, D.; Armas Padilla, M.; Bahramian, A.; Castro Segura, N.; Fogantini, F. A.; Heinke, C. O.; Maccarone, T.; Maitra, D.; Miller-Jones, J. C. A.; Muñoz-Darias, T.; Ozbey Arabaci, M.; Russell, D. M.; Shaw, A. W.; Sivakoff, G.; Tetarenko, A. J.; Vincentelli, F. Wijnands, R.

Abstract:

Swift J1858.6-0814 is a transient neutron star X-ray binary discovered in 2018 October. Multiwavelength follow-up observations across the electromagnetic spectrum revealed many interesting properties, such as erratic flaring on minute time-scales and evidence for wind outflows at both X-ray and optical wavelengths, strong and variable local absorption, and an anomalously hard X-ray spectrum. Here, we report on a detailed radio observing campaign consisting of one observation at 5.5/9 GHz with the Australia Telescope Compact Array, and nine observations at 4.5/7.5 GHz with the Karl G. Jansky Very Large Array. A radio counterpart with a flat to inverted radio spectrum is detected in all observations, consistent with a compact jet being launched from the system. Swift J1858.6-0814 is highly variable at radio wavelengths in most observations, showing significant variability when imaged on 3-to-5-min time-scales and changing up to factors of 8 within 20 min. The periods of brightest radio emission are not associated with steep radio spectra, implying they do not originate from the launching of discrete ejecta. We find that the radio variability is similarly unlikely to have a geometric origin, be due to scintillation, or be causally related to the observed X-ray flaring. Instead, we find that it is consistent with being driven by variations in the accretion flow propagating down the compact jet. We compare the radio properties of SwiftJ1858.6-0814 with those of Eddington-limited X-ray binaries with similar X-ray and optical characteristics, but fail to find a match in radio variability, spectrum, and luminosity.

Evidence for Late-stage Eruptive Mass Loss in the Progenitor to SN2018gep, a Broad-lined Ic Supernova: Pre-explosion Emission and a Rapidly Rising Luminous Transient

The Astrophysical Journal American Astronomical Society 887:2 (2019) 169

Authors:

Anna YQ Ho, Daniel A Goldstein, Steve Schulze, David K Khatami, Daniel A Perley, Mattias Ergon, Avishay Gal-Yam, Alessandra Corsi, Igor Andreoni, Cristina Barbarino, Eric C Bellm, Nadia Blagorodnova, Joe S Bright, E Burns, S Bradley Cenko, Virginia Cunningham, Kishalay De, Richard Dekany, Alison Dugas, Rob P Fender, Claes Fransson, Christoffer Fremling, Adam Goldstein, Matthew J Graham, David Hale, Assaf Horesh, Tiara Hung, Mansi M Kasliwal, N Paul M Kuin, SR Kulkarni, Thomas Kupfer, Ragnhild Lunnan, Frank J Masci, Chow-Choong Ngeow, Peter E Nugent, Eran O Ofek, Maria T Patterson, Glen Petitpas, Ben Rusholme, Hanna Sai, Itai Sfaradi, David L Shupe, Jesper Sollerman, Maayane T Soumagnac, Yutaro Tachibana, Francesco Taddia, Richard Walters, Xiaofeng Wang, Yuhan Yao, Xinhan Zhang

Non-Gaussianity constraints using future radio continuum surveys and the multitracer technique

Monthly Notices of the Royal Astronomical Society Oxford University Press 492:1 (2019) 1513-1522

Authors:

Zahra Gomes, Stefano Camera, Matthew Jarvis, Catherine Hale, José Fonseca

Abstract:

Tighter constraints on measurements of primordial non-Gaussianity (PNG) will allow the differentiation of inflationary scenarios. The cosmic microwave background bispectrum – the standard method of measuring the local non-Gaussianity – is limited by cosmic variance. Therefore, it is sensible to investigate measurements of non-Gaussianity using the large-scale structure. This can be done by investigating the effects of non-Gaussianity on the power spectrum on large scales. In this study, we forecast the constraints on the local PNG parameter fNL that can be obtained with future radio surveys. We utilize the multitracer method that reduces the effect of cosmic variance and takes advantage of the multiple radio galaxy populations that are differently biased tracers of the same underlying dark matter distribution. Improvements on previous work include the use of observational bias and halo mass estimates, updated simulations, and realistic photometric redshift expectations, thus producing more realistic forecasts. Combinations of Square Kilometre Array simulations and radio observations were used as well as different redshift ranges and redshift bin sizes. It was found that in the most realistic case the 1σ error on fNL falls within the range 4.07–6.58, rivalling the tightest constraints currently available.