Dr Peter Hatfield

Abstract:

The accurate estimation of photometric redshifts is crucial to many upcoming galaxy surveys, for example, the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST). Almost all Rubin extragalactic and cosmological science requires accurate and precise calculation of photometric redshifts; many diverse approaches to this problem are currently in the process of being developed, validated, and tested. In this work, we use the photometric redshift code GPz to examine two realistically complex training set imperfections scenarios for machine learning based photometric redshift calculation: (i) where the spectroscopic training set has a very different distribution in color–magnitude space to the test set, and (ii) where the effect of emission line confusion causes a fraction of the training spectroscopic sample to not have the true redshift. By evaluating the sensitivity of GPz to a range of increasingly severe imperfections, with a range of metrics (both of photo-z point estimates as well as posterior probability distribution functions, PDFs), we quantify the degree to which predictions get worse with higher degrees of degradation. In particular, we find that there is a substantial drop-off in photo-z quality when line-confusion goes above ∼1%, and sample incompleteness below a redshift of 1.5, for an experimental setup using data from the Buzzard Flock synthetic sky catalogs.

Abstract:

In this paper we present photometric redshifts for 2.7 million galaxies in the XMM-LSS and COSMOS fields, both with rich optical and near-infrared data from VISTA and HyperSuprimeCam. Both template fitting (using galaxy and Active Galactic Nuclei templates within LePhare) and machine learning (using GPz) methods are run on the aperture photometry of sources selected in the Ks-band. The resulting predictions are then combined using a Hierarchical Bayesian model, to produce consensus photometric redshift point estimates and probability distribution functions that outperform each method individually. Our point estimates have a root mean square error of ∼0.08 − 0.09, and an outlier fraction of ∼3 − 4 percent when compared to spectroscopic redshifts. We also compare our results to the COSMOS2020 photometric redshifts, which contains fewer sources, but had access to a larger number of bands and greater wavelength coverage, finding that comparable photo-z quality can be achieved (for bright and intermediate luminosity sources where a direct comparison can be made). Our resulting redshifts represent the most accurate set of photometric redshifts (for a catalogue this large) for these deep multi-square degree multi-wavelength fields to date.

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 emulates simulations 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.