We propose a fast and accurate methodology for prediction of hydrodynamic gas distributions from Dark Matter N-body simulations. We perform fast emulations of 3D gas density cubes from the Dark Matter counterparts, comparing different Convolutional Neural Network architectures. Our method achieves an accuracy above 95% for the pixel density contrast distributions within the middle regions of the density contrast domain, and 98% accuracy in the matter power spectrum throughout the entire k domain. Moreover, our method provides a gain of 4 orders of magnitude in CPU run times compared to running the full hydrodynamic N-body simulation on a slightly newer computer system. Most importantly, our methodology provides a scalable and generalizable approach to the problem of N-body emulation, with the potential to be applied to simulations of arbitrary sizes and various scalar quantities.

Application-specific hardware acceleration of computation-intensive kernels can often provide significant performance and power efficiency improvements over general-purpose software, but it is difficult and costly to incorporate them into existing software systems. Designing hardware accelerators and modifying legacy software to incorporate them are already complex tasks. Furthermore, identifying a suitable kernel for acceleration is complicated by the PCIe-attached architecture of commodity FPGA cards, meaning bandwidth and latency overhead must be considered for kernel selection. For example, small blocking kernels may not benefit from acceleration due to PCIe latency. As a remedy, we present FarSlayer, a high-level source-to-source compiler for end-to-end acceleration of legacy software. FarSlayer analyzes existing software code and emits an accelerated version of it, where the kernel is automatically selected considering data movement over PCIe. Specifically, FarSlayer identifies kernels which can be called asynchronously to hide the PCIe latency, while also having a high operational intensity for low bandwidth requirements. The entire process is automatic, meaning the programmer does not necessarily need to understand the existing code, or reason about hardware development. We demonstrate FarSlayer on multiple existing scientific computing software systems, and demonstrate it can automatically achieve significant performance improvements.

Computations of cosmological properties, such as the density perturbation variance,$\sigma$, are computationally expensive. In this work we propose the application of Symbolic Regression to generate analytical expressions to approximate these quantities. We created simulated data for$\sigma$using a Boltzmann solver, CAMB. These simulations cover seven parameters; the five cosmological parameters and the redshift and mass of dark matter halos. We then apply a Symbolic Regression engine, TuringBot, to this simulated data and obtain an analytical equation to approximate$\sigma$. The resulting mathematical expression has a mean accuracy of ≈ 98.96% over the entire domain, and is five orders of magnitude faster than using simulated data from CAMB, demonstrating the applicability of Symbolic Regression to accelerate cosmological inference.

In this work, we propose a prototype for a scalable software infrastructure to enable the retrieval and analysis of galaxy images from public astronomical archives. We implement and deploy our prototype to estimate the position angles of$6 \times 10^{6}$galaxies in multiple optical to near-infrared wavelengths. The position angle determination is performed using Radon transforms, and thus it is obtained in a fully non-parametric way, reducing possible model-dependent biases. Finally, these measurements can be adopted to further reveal dark matter signatures and constrain important cosmological parameters of the models we use to explain our Universe.

In this work, we propose an efficient methodology for emulating hydrodynamic gas structures in Dark Matter Nbody simulations. We perform fast emulations of 3D gas density cubes from the Dark Matter counterparts, comparing different machine learning approaches, namely Principal Component Analysis + Random Forest ($\text{PCA}+\text{RF}$) and a Convolutional Neural Network (CNN). The method provides a gain of 5 orders of magnitude in CPU run times compared to running the full hydrodynamic N-body simulation in the same computer system. Using$\text{PCA}+\text{RF}$, the method achieves 98% accuracy in the matter power spectrum compared to the full hydrodynamic simulation throughout most of the k domain$(k < 1.0\ Mpc\ h^{-1})$. Finally, CNNs offer increased accuracy (∼ 99%), showing a striking improvement in performance at the extremes of the k domain.

Time series from real data measurements are often noisy, under-sampled, irregularly sampled, and inconsistent across long-term measurements. Typically, in analyzing these time series, particularly within astronomy, it is common to use estimators such as sample entropy and multi-scale entropy that require interpolation to avoid irregular sampling. In this work, we analyze and consider a new entropy estimator that combines permutations, Fourier Coefficients, and Wasserstein distances to address the concern of irregularly sampled data.

Gravitationally lensed quasars are important objects in astronomy for probing the universe. Unfortunately, these objects are exceedingly rare, occurring only for only ∼ 1/10 000 quasars. The challenge is to find these lensed quasars amongst large astronomical data sets. In contrast to previous attempts, which have only made use of numeric data, we perform semi-supervised classification based on images of quasars. These images are low resolution and noisy, but are enough for experienced astronomers to perform classification. Using virtual adversarial training to take advantage of millions of unlabelled images, we develop a classifier which achieves an F1 score of 0.49 — an extremely impressive result in this domain. Predictions made by this classifier are already being used to select candidates for telescopes around the world.

Quantum computers are becoming a reality. The advantage of quantum computing is that it has the potential to solve computationally complex problems in a fixed amount time, independent of the size of the problem. However, the kinds of problems for which these computers are a good fit, and the ways to express those problems, are substantially different from the kinds of problems and expressions used in classical computing. Quantum annealers, in particular, are currently the most promising and available quantum computing devices in the short term. However, they are also the most foreign compared to classical programs, as they require a different kind of computational thinking. In order to ease the transition into this new world of quantum computing, we present a novel quantum approach to a well-known software problem: code clone detection. We express code clone detection as a subgraph isomorphism problem that is mapped into a quadratic optimization problem, and solve it using a DWave quantum annealing computer. We developed a quantum annealing algorithm that compares Abstract Syntax Trees (AST) and reports an energy value that indicates how similar they are.

In this work, we propose a resolution enhancement methodology for discretized scalar fields and apply it to cosmological N-body simulations. We upscale Dark Matter density 3D Eulerian cubes using supervised machine learning and Principal Component Analysis. Once the low-resolution simulation is performed, our methodology doubles its resolution gaining three orders of magnitude in CPU run times compared to performing the full high-resolution simulation in the same computational system. Moreover, we achieve 98% accuracy in the matter power spectrum compared to the full high-resolution simulation throughout most of the k domain$(k < 1.0\ Mpc\ h^{-1})$. Finally, the proposed approach is agnostic to the nature of the simulation field. as long as it corresnonds to a 3D scalar field.

In this work, we analyze and compare two entropy estimators applied to random walk time series. We compare the robustness of multi-scale entropy and sample entropy for different regimes of signal-to-noise ratio. We also compare multi-scale entropy and sample entropy in the case of missing data when simple linear interpolation is adopted to fill the missing data points. In the case of the signal-to-noise comparison, we show by numerical simulations and present strong mathematical arguments that multi-scale entropy is a more resistant estimator to analyze time series. We also show that multi-scale entropy provides a more resistant and accurate estimate of entropy on random walk time series in the scenario of missing data, especially when completing missing data with linear interpolation.

The study of cosmological structure formation usually relies on computationally intensive N-body simulations, that evolve ensembles of particles assuming an underlying physical model. The diversity of physical assumptions and the number of parameters involved often limit the application of these techniques to only a few cases in the multi-dimensional spaces of cosmological parameters. Recently, supervised deep learning methods have been proposed to alleviate part of the computational overhead of N-body methods, but overheads are still high. In this work, we present a new method, based on Functional Principal Component Analysis (FPCA), that allows fast and accurate estimations of 3D N-body density fields and run-time gains of orders of magnitude when compared with traditional N-body simulations of the same resolution. Here we also compare FPCA to an approach of compressing the simulations through Principal Component Analysis (PCA) and using a single-layer Neural Network (NN) to perform the emulation. We show that FPCA can achieve slightly better accuracy than the PCA+NN method, short-cutting the need to apply supervised learning as an additional step in the emulation of cosmological simulations.

Consider a set of points P in space with at least some of the pairwise distances specified. Given this set P, consider the following three kinds of queries against a database D of points : (i) pure constellation query: find all sets S in D of size |P| that exactly match the pairwise distances within P up to an additive error ϵ; (ii) isotropic constellation queries: find all sets S in D of size |P| such that there exists some scale factor f for which the distances between pairs in S exactly match f times the distances between corresponding pairs of P up to an additive ϵ; (iii) non-isotropic constellation queries: find all sets S in D of size |P| such that there exists some scale factor f and for at least some pairs of points, a maximum stretch factor mi,j > 1 such that (f X mi,jXdist(pi, pj))+ϵ > dist(si,sj) > (f X dist(pi, pj)) - ϵ. Finding matches to such queries has applications to spatial data in astronomical, seismic, and any domain in which (approximate, scale-independent) geometrical matching is required. Answering the isotropic and non-isotropic queries is challenging because scale factors and stretch factors may take any of an infinite number of values. This paper proposes practically efficient sequential and distributed algorithms for pure, isotropic, and non-isotropic constellation queries. As far as we know, this is the first work to address isotropic and non-isotropic queries.