Huang et al (J Agric Biol Environ Stat, 2023,
https://doi.org/10.1007/s13253-022-00518-x
) a suite of statistical models for storage-efficient climate model emulation. In this discussion, I review ...and explore possibility of using machine learning methods, in particular, deep neural network (DNN)-based variational autoencoders (VAE) for the same task of spatio-temporal climate data compression. I discuss the pros and cons of the statistical and the machine learning paradigms.
Spatial process models for analyzing geostatistical data entail computations that become prohibitive as the number of spatial locations become large. This article develops a class of highly scalable ...nearest-neighbor Gaussian process (NNGP) models to provide fully model-based inference for large geostatistical datasets. We establish that the NNGP is a well-defined spatial process providing legitimate finite-dimensional Gaussian densities with sparse precision matrices. We embed the NNGP as a sparsity-inducing prior within a rich hierarchical modeling framework and outline how computationally efficient Markov chain Monte Carlo (MCMC) algorithms can be executed without storing or decomposing large matrices. The floating point operations (flops) per iteration of this algorithm is linear in the number of spatial locations, thereby rendering substantial scalability. We illustrate the computational and inferential benefits of the NNGP over competing methods using simulation studies and also analyze forest biomass from a massive U.S. Forest Inventory dataset at a scale that precludes alternative dimension-reducing methods. Supplementary materials for this article are available online.
The low‐latitude ionosphere is a dynamic region with a wide range of disturbances in temporal and spatial scales. The Giant Metrewave Radio Telescope (GMRT) situated in the low‐latitude region has ...demonstrated its ability to detect various ionospheric phenomena. It can detect total electron content (TEC) variation with precision of 10−3 TECU and also can measure TEC gradient with an accuracy of about 7 × 10−4 TECU km−1. This paper describes the spectral analysis of previously calculated TEC gradient measurements and validates them by comparing their properties using two bands. The analysis tracked individual waves associated with medium‐scale traveling ionospheric disturbances (MSTIDs) and smaller waves down to wavelengths of ∼10 km. The ionosphere is found to have unanticipated changes during sunrise hours, with waves changed propagation direction as the sun approached the zenith. Equatorial spread F disturbances during sunrise hours is observed, along with smaller structures moving in the same direction.
Plain Language Summary
The Earth's ionosphere can limit exploring sub‐GHz frequencies of the sky and introduces an extra phase term that is difficult to calibrate. The same calibration data can be used to study the Earth's ionosphere more precisely than conventional probes. Radio interferometry is a technique for studying astronomical sources and Earth's ionosphere by measuring the spatial coherence function of multiple elements. The GMRT is a unique instrument for exploring the equatorial ionosphere region. This study used dual‐band observations of a bright radio source with the GMRT to explore the Equatorial Ionization Anomaly region. The GMRT can detect variations in total electron content and measure TEC gradient with high accuracy. Spectral analysis was performed on TEC gradient measurements to track individual waves associated with medium scales traveling ionospheric disturbances and smaller waves up to wavelengths of about ∼10 km. The results showed unexpected changes in the ionosphere during sunrise hours and observed large plasma irregularities and smaller structures moving in the same direction.
Key Points
Giant Metrewave Radio Telescope (GMRT) can demonstrate an order of magnitude better sensitivity than GNSS‐based TEC measurements in characterizing ionospheric fluctuations
The spectral analysis technique used with GMRT can detect multiple MSTIDs and smaller‐scale structures simultaneously
GMRT can detect ionospheric variations as small as 10 km. The study also showed waves changing direction unexpectedly during sunrise time
With the advent of high throughput genetic data, there have been attempts to estimate heritability from genome-wide SNP data on a cohort of distantly related individuals using linear mixed model ...(LMM). Fitting such an LMM in a large scale cohort study, however, is tremendously challenging due to its high dimensional linear algebraic operations. In this paper, we propose a new method named PredLMM approximating the aforementioned LMM motivated by the concepts of genetic coalescence and Gaussian predictive process. PredLMM has substantially better computational complexity than most of the existing LMM based methods and thus, provides a fast alternative for estimating heritability in large scale cohort studies. Theoretically, we show that under a model of genetic coalescence, the limiting form of our approximation is the celebrated predictive process approximation of large Gaussian process likelihoods that has well-established accuracy standards. We illustrate our approach with extensive simulation studies and use it to estimate the heritability of multiple quantitative traits from the UK Biobank cohort.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
The Gaussian process is an indispensable tool for spatial data analysts. The onset of the “big data” era, however, has lead to the traditional Gaussian process being computationally infeasible for ...modern spatial data. As such, various alternatives to the full Gaussian process that are more amenable to handling big spatial data have been proposed. These modern methods often exploit low-rank structures and/or multi-core and multi-threaded computing environments to facilitate computation. This study provides, first, an introductory overview of several methods for analyzing large spatial data. Second, this study describes the results of a predictive competition among the described methods as implemented by different groups with strong expertise in the methodology. Specifically, each research group was provided with two training datasets (one simulated and one observed) along with a set of prediction locations. Each group then wrote their own implementation of their method to produce predictions at the given location and each was subsequently run on a common computing environment. The methods were then compared in terms of various predictive diagnostics.
ABSTRACT
The redshifted 21-cm signal of neutral hydrogen is a promising probe into the period of evolution of our Universe when the first stars were formed (Cosmic Dawn), to the period where the ...entire Universe changed its state from being completely neutral to completely ionized (Reionization). The most striking feature of this line of neutral hydrogen is that it can be observed across an entire frequency range as a sky-averaged continuous signature, or its fluctuations can be measured using an interferometer. However, the 21-cm signal is very faint and is dominated by a much brighter Galactic and extragalactic foregrounds, making it an observational challenge. We have used different physical models to simulate various realizations of the 21-cm global signals, including an excess radio background to match the amplitude of the Experiment to Detect the Global EoR Signature (EDGES) 21-cm signal. First, we have used an artificial neural network (ANN) to extract the astrophysical parameters from these simulated data sets. Then, mock observations were generated by adding a physically motivated foreground model and an ANN was used to extract the astrophysical parameters from such data. The R2 score of our predictions from the mock observations is in the range of 0.65–0.89. We have used this ANN to predict the signal parameters giving the EDGES data as the input. We find that the reconstructed signal closely mimics the amplitude of the reported detection. The recovered parameters can be used to infer the physical state of the gas at high redshifts.
ABSTRACT
Detection of the H i 21-cm power spectrum is one of the key science drivers of several ongoing and upcoming low-frequency radio interferometers. However, the major challenge in such ...observations come from bright foregrounds, whose accurate removal or avoidance is key to the success of these experiments. In this work, we demonstrate the use of artificial neural networks (ANNs) to extract the H i 21-cm power spectrum from synthetic data sets and extract the reionization parameters from the H i 21-cm power spectrum. For the first time, using a suite of simulations, we present an ANN-based framework capable of extracting the H i signal power spectrum directly from the total observed sky power spectrum (which contains the 21-cm signal, along with the foregrounds and effects of the instrument). We have used a combination of two ANNs sequentially. In the first step, ANN1 predicts the 21-cm power spectrum directly from foreground corrupted synthetic data sets. In the second step, ANN2 predicts the reionization parameters from the predicted H i power spectra from ANN1. The two-step ANN framework can be used as an alternative method to extract the 21-cm power spectrum and the reionization parameters directly from foreground dominated data sets. Our ANN-based framework is trained at a redshift of 9.01, and for $\boldsymbol {k}$ modes in the range, $\rm {0.17\lt {\boldsymbol {k}}\lt 0.37~Mpc^{-1}}$. We have tested the network’s performance with mock data sets corrupted with thermal noise corresponding to 1080 h of observations of the SKA-1 LOW and HERA. We have recovered the H i power spectra from foreground dominated synthetic data sets, with an accuracy of $\approx 95{\!-\!}99{{\ \rm per\ cent}}$. We have achieved an accuracy of $\approx ~81{\!-\!}90{{\ \rm per\ cent}}$ and $\approx ~50{\!-\!}60{{\ \rm per\ cent}}$ for the predicted reionization parameters, for test sets corrupted with thermal noise corresponding to the SKA-1 LOW and HERA, respectively.
This paper describes and illustrates functionality of the spNNGP R package. The package provides a suite of spatial regression models for Gaussian and non-Gaussian pointreferenced outcomes that are ...spatially indexed. The package implements several Markov chain Monte Carlo (MCMC) and MCMC-free nearest neighbor Gaussian process (NNGP) models for inference about large spatial data. Non-Gaussian outcomes are modeled using a NNGP Pólya-Gamma latent variable. OpenMP parallelization options are provided to take advantage of multiprocessor systems. Package features are illustrated using simulated and real data sets.
The study of solar coronal holes (CHs) is important in the understanding of solar physics and the prediction of space weather events, which have significant impact on space-based instruments, ...communication and navigation systems. With the availability of the multi-wavelength Atmospheric Imaging Assembly (AIA) instrument on board Solar Dynamics Observatory (SDO) satellite, a large volume of high-resolution solar images are produced continuously. Proper detection of CHs from AIA images is an important issue and recently, a few contour and machine learning-based techniques are found to be promising for such purpose. However, accuracy, time complexity and the requirement of human intervention are some of the critical issues with such methods. In this paper, to address these challenging issues, two contour-based approaches are developed, namely i) the Hough transformed simulated parameterized online region-based active contour method (POR-ACM) and ii) fast fuzzy c-means clustering followed by Hough transformed simulated static contour method (FFCM-SCM). The major issues that are addressed here are automated initialization of contour, reducing time complexity and avoidance of non-coronal hole inside a coronal hole region during contour evolution. The proposed techniques have been tested on three benchmark solar disk images and compared with the existing active contour without edge- (ACWE) based method and fuzzy energy-based dual contour method (FEDCM) of CHs segmentation. The results indicate the capability of the proposed techniques in detection and extraction of CHs in solar disk image with higher accuracy and reduced time complexity.
Noise leads to the ambiguity in regions of interest detection by corrupting the pixel information and is a vital problem in image processing domain. A novel hybrid technique based on fuzzy filtering ...and fuzzy thresholding is proposed here to extract the object regions accurately in presence of Gaussian noises. The proposed method is automated, does not need any parameter tuning as well does not need prior knowledge of the image or noise. An asymmetrical triangular fuzzy filter with median center coupled with a thresholding based on fuzziness minimization technique are implemented for this purpose. The fuzzy thresholding technique helps to classify the pixels with low signal-to-noise ratio (SNR) caused either due to noise or by the application of noise removal process. The proposed technique is applied in benchmark images corrupted by noises and are compared with some of the popular algorithms of object detection. The results indicate that the proposed method has superior performance in terms of peak signal-to-noise ratio (PSNR) and mean square error (MSE) value for images corrupted with Gaussian noises with standard deviation upto 1.5.