Day 1487

Monday.

1810.08430
An application of machine learning techniques to galaxy cluster mass estimation using the MACSIS simulations
Armitage, Kay, Barnes

ML techniques, in particular supervised regression algorithms, are a promising new way to use multiple observables to predict a cluster's mass or any other key features.  To investigate this approach, use the MACSIS sample of simulated Hydro galaxy clusters to train a variety of ML models, mimicking different datasets.  Find that compared to predicting the cluster mass from the sigma-M relation, the scatter in the predicted-to-true mass ratio is reduced by a factor of 4, from 0.130±0.004 dex ~ 35% to 0.031±0.001 dex ~ 7% when using the same, interloper contaminated, spectroscopic galaxy sample.  Interestingly, omitting line-of-sight galaxy velocities from the training set has no effect on the scatter the the galaxies are taken from within r_200c. Also train ML models to reproduce estimated masses derived from the mock X-ray and WL analyses.  While the WL masses can be recovered with a similar scatter to that when training on the true mass, the hydrostatic mass suffers from significantly higher scatter of ~0.13 dex ~35%.  Training models using dark matter only simulations does not significantly increase the scatter in predicted cluster mass compared to training on simulated clusters with hydrodynamics.  In summary, find ML techniques to offer a powerful method to predict masses for large samples of clusters, a vital requirement for cosmological analysis with future surveys.


1810.08513
Indirect detection of extrasolar planets via Astrometry
Butler, Matthews

Radio wavelength astrometry of stars and other objects has a long and productive history. The use of that technique to determine whether stars have planets around them would cover a nearly unique part of the parameter space for detection of those systems.  Namely, astrometric observations are most sensitive to systems with large planets in moderately wide orbits (a few to ~10 AU), because it is those systems that produce large reflex motion of the star, in a short enough measurement period (years to tens of years).  In addition, astrometric observations are most sensitive to systems with planets in close orbits (less than ~1 AU), which are nearly edge-on.  Describe here, using the Hipparcos and Gaia star catalogs, how ngVLA could use this technique on hundreds of stars, some tens of which are solar analogs, to determine whether these stars have planets orbiting them.