Search

We consider hate speech detection through keyword spotting on radio broadcasts. One approach is to build an automatic speech recognition (ASR) system for the target low-resource language. We compare this to using acoustic word embedding (AWE) models that map speech segments to a space where matching words have similar vectors. We specifically use a multilingual AWE model trained on labelled data from well-resourced languages to spot keywords in data in the unseen target language. In contrast to ASR, the AWE approach only requires a few keyword exemplars. In controlled experiments on Wolof and Swahili where training and test data are from the same domain, an ASR model trained on just five minutes of data outperforms the AWE approach. But in an in-the-wild test on Swahili radio broadcasts with actual hate speech keywords, the AWE model (using one minute of template data) is more robust, giving similar performance to an ASR system trained on 30 hours of labelled data., Accepted to Interspeech 2023

Radio interferometry calibration and Radio Frequency Interference (RFI) removal are usually done separately. Here we show that jointly modelling the antenna gains and RFI has significant benefits when the RFI follows precise trajectories, such as for satellites. One surprising benefit is improved calibration solutions, by leveraging the RFI signal itself. We present tabascal (TrAjectory BAsed RFI Subtraction and CALibration), a new algorithm that jointly models the RFI and calibration parameters in visibilities. We test tabascal on simulated MeerKAT calibration observations contaminated by satellite-based RFI. We obtain gain estimates that are both unbiased and up to an order of magnitude better constrained compared to uncontaminated data. When combined with an ad hoc RFI subtraction scheme, tabascal solutions can be further applied to an adjacent target observation: 5 minutes of calibration data results in an image with about a third the noise achieved when using flagging alone. The recovered flux distribution of RFI subtracted data was on par with uncontaminated data. In contrast, RFI flagging alone resulted in a higher detection threshold and consistent underestimation of source fluxes. For a mean RFI amplitude of 17 Jy, using RFI subtraction leads to less than 1% loss of data compared to 75% data loss from an ideal $3\sigma$ flagging algorithm, a very significant increase in data available for science analysis. Although we have examined the case of satellite RFI, tabascal should work for any RFI moving on parameterizable trajectories, relative to the phase centre, such as planes and/or objects fixed to the ground., Comment: 21 pages, 17 figures, 4 tables

Active learning aims to deliver maximum benefit when resources are scarce. We use COMET-QE, a reference-free evaluation metric, to select sentences for low-resource neural machine translation. Using Swahili, Kinyarwanda and Spanish for our experiments, we show that COMET-QE significantly outperforms two variants of Round Trip Translation Likelihood (RTTL) and random sentence selection by up to 5 BLEU points for 20k sentences selected by Active Learning on a 30k baseline. This suggests that COMET-QE is a powerful tool for sentence selection in the very low-resource limit., Accepted to Findings of EMNLP 2022

The next generation of telescopes such as the SKA and the Rubin Observatory will produce enormous data sets, requiring automated anomaly detection to enable scientific discovery. Here, we present an overview and friendly user guide to the Astronomaly framework for active anomaly detection in astronomical data. Astronomaly uses active learning to combine the raw processing power of machine learning with the intuition and experience of a human user, enabling personalised recommendations of interesting anomalies. It makes use of a Python backend to perform data processing, feature extraction and machine learning to detect anomalous objects; and a JavaScript frontend to allow interaction with the data, labelling of interesting anomalous and active learning. Astronomaly is designed to be modular, extendable and run on almost any type of astronomical data. In this paper, we detail the structure of the Astronomaly code and provide guidelines for basic usage., Comment: 10 pages, 3 figures, to be published in Proc. ADASS XXXI (2021), coincides with v1.2 release of Astronomaly (https://github.com/MichelleLochner/astronomaly)

Radio Frequency Interference (RFI) is unwanted noise that swamps the desired astronomical signal. Radio astronomers have always had to deal with RFI detection and excision around telescope sites, but little has been done to understand the full scope, nature and evolution of RFI in a unified way. We undertake this for the MeerKAT array using a probabilistic multidimensional framework approach focussing on UHF-band and L-band data. In the UHF- band, RFI is dominated by the allocated Global System for Mobile (GSM) Communications, flight Distance Measuring Equipment (DME), and UHF-TV bands. The L-band suffers from known RFI sources such as DMEs, GSM, and the Global Positioning System (GPS) satellites. In the "clean" MeerKAT band, we noticed the RFI occupancy changing with time and direction for both the L-band and UHF band. For example, we saw a significant increase (300% increase) in the fraction of L-band flagged data in November 2018 compared to June 2018. This increase seems to correlate with construction activity on site. In the UHF-band, we found that the early morning is least impacted by RFI and other outliers. We also found a dramatic decrease in DME RFI during the hard lockdown due to the COVID-19 pandemic. The work presented here allows us to characterise the evolution of RFI at the MeerKAT site. Any observatory can adopt it to understand the behaviour of RFI within its surroundings.

Anomaly detection algorithms are typically applied to static, unchanging, data features hand-crafted by the user. But how does a user systematically craft good features for anomalies that have never been seen? Here we couple deep learning with active learning -- in which an Oracle iteratively labels small amounts of data selected algorithmically over a series of rounds -- to automatically and dynamically improve the data features for efficient outlier detection. This approach, AHUNT, shows excellent performance on MNIST, CIFAR10, and Galaxy-DESI data, significantly outperforming both standard anomaly detection and active learning algorithms with static feature spaces. Beyond improved performance, AHUNT also allows the number of anomaly classes to grow organically in response to Oracle's evaluations. Extensive ablation studies explore the impact of Oracle question selection strategy and loss function on performance. We illustrate how the dynamic anomaly class taxonomy represents another step towards fully personalized rankings of different anomaly classes that reflect a user's interests, allowing the algorithm to learn to ignore statistically significant but uninteresting outliers (e.g., noise). This should prove useful in the era of massive astronomical datasets serving diverse sets of users who can only review a tiny subset of the incoming data., Comment: 10 pages, 7 figures

Low-resource African languages have not fully benefited from the progress in neural machine translation because of a lack of data. Motivated by this challenge we compare zero-shot learning, transfer learning and multilingual learning on three Bantu languages (Shona, isiXhosa and isiZulu) and English. Our main target is English-to-isiZulu translation for which we have just 30,000 sentence pairs, 28% of the average size of our other corpora. We show the importance of language similarity on the performance of English-to-isiZulu transfer learning based on English-to-isiXhosa and English-to-Shona parent models whose BLEU scores differ by 5.2. We then demonstrate that multilingual learning surpasses both transfer learning and zero-shot learning on our dataset, with BLEU score improvements relative to the baseline English-to-isiZulu model of 9.9, 6.1 and 2.0 respectively. Our best model also improves the previous SOTA BLEU score by more than 10.

Deep facial expression recognition faces two challenges that both stem from the large number of trainable parameters: long training times and a lack of interpretability. We propose a novel method based on evolutionary algorithms, that deals with both challenges by massively reducing the number of trainable parameters, whilst simultaneously retaining classification performance, and in some cases achieving superior performance. We are robustly able to reduce the number of parameters on average by 95% (e.g. from 2M to 100k parameters) with no loss in classification accuracy. The algorithm learns to choose small patches from the image, relative to the nose, which carry the most important information about emotion, and which coincide with typical human choices of important features. Our work implements a novel form attention and shows that evolutionary algorithms are a valuable addition to machine learning in the deep learning era, both for reducing the number of parameters for facial expression recognition and for providing interpretable features that can help reduce bias., Conference of the South African Institute of Computer Scientists and Information Technologists 2020

The Infection Fatality Rate (IFR) for COVID-19 is a poorly known, yet crucial, aspect of the disease. Counting only current deaths in a region and assuming everyone in that region is infected provides an absolute lower bound on the IFR. Using this estimator for New York City, Lombardy and Madrid yields strong bounds on the average IFR in overwhelmed health systems. Their combined 35, 152 deaths implies an IFR ≥ 0.14% averaged over 25.1 million people. This is the best-case scenario and conclusively demonstrates that COVID-19 is more deadly than influenza. The actual value of the average COVID-19 IFR is likely to be higher than this bound.

Radio Frequency Interference (RFI) has historically plagued radio astronomy, worsening with the rapid spread of electronics and increasing telescope sensitivity. We present a multi-dimensional probabilistic framework for characterising the RFI environment around a radio astronomy site that uses automatically flagged data from the array itself. We illustrate the framework using about 1500 hours of commissioning data from the MeerKAT radio telescope; producing a 6-dimensional array that yields both average RFI occupancy as well as confidence intervals around the mean as a function of key variables (frequency, direction, baseline, time). Our results provide the first detailed view of the MeerKAT RFI environment at high sensitivity as a function of direction, frequency, time of day and baseline. They allow us to track the historical evolution of the RFI and to quantify fluctuations which can be used for alerting on new RFI. As expected we find the major RFI contributors for MeerKAT site are from Global Positioning System (GPS) satellites, flight Distance Measurement Equipment (DME) and the Global System for Mobile (GSM) Communications. Beyond characterising RFI environments our approach allows observers access to the prior probability of RFI in any combination of tracked variables, allowing for more efficient observation planning and data excision.

Statistical uncertainties are rarely incorporated in machine learning algorithms, especially for anomaly detection. Here we present the Bayesian Anomaly Detection And Classification (BADAC) formalism, which provides a unified statistical approach to classification and anomaly detection within a hierarchical Bayesian framework. BADAC deals with uncertainties by marginalising over the unknown, true, value of the data. Using simulated data with Gaussian noise, BADAC is shown to be superior to standard algorithms in both classification and anomaly detection performance in the presence of uncertainties, though with significantly increased computational cost. Additionally, BADAC provides well-calibrated classification probabilities, valuable for use in scientific pipelines. We show that BADAC can work in online mode and is fairly robust to model errors, which can be diagnosed through model-selection methods. In addition it can perform unsupervised new class detection and can naturally be extended to search for anomalous subsets of data. BADAC is therefore ideal where computational cost is not a limiting factor and statistical rigour is important. We discuss approximations to speed up BADAC, such as the use of Gaussian processes, and finally introduce a new metric, the Rank-Weighted Score (RWS), that is particularly suited to evaluating the ability of algorithms to detect anomalies., Comment: 29 pages, 13 figures, Demo available: https://github.com/ethyroberts/BADAC

New telescopes like the Square Kilometre Array (SKA) will push into a new sensitivity regime and expose systematics, such as direction-dependent effects, that could previously be ignored. Current methods for handling such systematics rely on alternating best estimates of instrumental calibration and models of the underlying sky, which can lead to inadequate uncertainty estimates and biased results because any correlations between parameters are ignored. These deconvolution algorithms produce a single image that is assumed to be a true representation of the sky, when in fact it is just one realization of an infinite ensemble of images compatible with the noise in the data. In contrast, here we report a Bayesian formalism that simultaneously infers both systematics and science. Our technique, Bayesian Inference for Radio Observations (BIRO), determines all parameters directly from the raw data, bypassing image-making entirely, by sampling from the joint posterior probability distribution. This enables it to derive both correlations and accurate uncertainties, making use of the flexible software MEQTREES to model the sky and telescope simultaneously. We demonstrate BIRO with two simulated sets of Westerbork Synthesis Radio Telescope data sets. In the first, we perform joint estimates of 103 scientific (flux densities of sources) and instrumental (pointing errors, beamwidth and noise) parameters. In the second example, we perform source separation with BIRO. Using the Bayesian evidence, we can accurately select between a single point source, two point sources and an extended Gaussian source, allowing for 'super-resolution' on scales much smaller than the synthesized beam., Comment: Published in MNRAS. See https://vimeo.com/117391380 for a video of MultiNest converging to the correct source model

We outline a new method to compute the Bayes Factor for model selection which bypasses the Bayesian Evidence. Our method combines multiple models into a single, nested, Supermodel using one or more hyperparameters. Since the models are now nested the Bayes Factors between the models can be efficiently computed using the Savage-Dickey Density Ratio (SDDR). In this way model selection becomes a problem of parameter estimation. We consider two ways of constructing the supermodel in detail: one based on combined models, and a second based on combined likelihoods. We report on these two approaches for a Gaussian linear model for which the Bayesian evidence can be calculated analytically and a toy nonlinear problem. Unlike the combined model approach, where a standard Monte Carlo Markov Chain (MCMC) struggles, the combined-likelihood approach fares much better in providing a reliable estimate of the log-Bayes Factor. This scheme potentially opens the way to computationally efficient ways to compute Bayes Factors in high dimensions that exploit the good scaling properties of MCMC, as compared to methods such as nested sampling that fail for high dimensions., Comment: 24 pages, 11 Figures

Classifying transients based on the multi band light curves is a challenging but crucial problem in the era of GAIA and LSST since the sheer volume of transients will make spectroscopic classification unfeasible. Here we present a nonparametric classier that uses the transient's light curve measurements to predict its class given training data. It implements two novel components: the first is the use of the BAGIDIS wavelet methodology - a method of characterizing functional data using hierarchical wavelet coefficients. The second novelty is the introduction of a ranked probability classier on the wavelet coefficients which handles both the variable size of measurement errors (heteroscedasticity) of the data in addition to the potential non-representativity of the training set. The ranked classier is simple and quick to implement while a major advantage of the BAGIDIS wavelets is that they are translation invariant, hence they do not need the light curves to be aligned to extract features. Further, the BAGIDIS methodology is nonparametric so it can be used for blind searches for new objects. We demonstrate the effectiveness of our ranked wavelet classier against the well- tested Supernova Photometric Classification Challenge (SNPCC) dataset in which the challenge is to correctly classify light curves as belonging to Type Ia or non- Ia supernovae. We train our ranked probability classier on the spectroscopically- confirmed subsample (which is not representative) and show that it gives good results for all supernova with observed light curve timespans greater than 100 days (roughly 55% of the dataset). For such data, we obtain a Ia efficiency (recall) of 80.5% and a purity (precision) of 82.4% yielding a highly competitive score of 0.49 averaged over all redshifts whilst implementing a truly \model-blind" approach to supernova classification. The classier compares favourably to standard algorithms such as k- Nearest Neighbours (kNN) and Support Vector Machines (SVM), which obtain scores of 0.45 and 0.41 respectively. Consequently this approach may be particularly suitable for the classification of astronomical transients in the era of large synoptic sky surveys.

There are two redshifts in cosmology: $z_{obs}$, the observed redshift computed via spectral lines, and the model redshift, $z$, defined by the effective FLRW scale factor. In general these do not coincide. We place observational constraints on the allowed distortions of $z$ away from $z_{obs}$ - a possibility we dub redshift remapping. Remapping is degenerate with cosmic dynamics for either $d_L(z)$ or $H(z)$ observations alone: for example, the simple remapping $z = \alpha_1 z_{obs} +\alpha_2 z_{obs}^2$ allows a decelerating Einstein de Sitter universe to fit the observed supernova Hubble diagram as successfully as $\Lambda$CDM, highlighting that supernova data alone cannot prove that the universe is accelerating. We show however, that redshift remapping leads to apparent violations of cosmic distance duality that can be used to detect its presence even when neither a specific theory of gravity nor the Copernican Principle are assumed. Combining current data sets favours acceleration but does not yet rule out redshift remapping as an alternative to dark energy. Future surveys, however, will provide exquisite constraints on remapping and any models -- such as backreaction -- that predict it., Comment: 12 pages 13 figures. Significantly generalised to discuss the case where distances and Hubble rate are arbitrarily deformed and General Relativity is not assumed

We introduce Bayesian Estimation Applied to Multiple Species (BEAMS), an algorithm designed to deal with parameter estimation when using contaminated data. We present the algorithm and demonstrate how it works with the help of a Gaussian simulation. We then apply it to supernova data from the Sloan Digital Sky Survey (SDSS), showing how the resulting confidence contours of the cosmological parameters shrink significantly., 23 pages, 9 figures. Chapter 4 in "Astrostatistical Challenges for the New Astronomy" (Joseph M. Hilbe, ed., Springer, New York, forthcoming in 2012), the inaugural volume for the Springer Series in Astrostatistics

The original formulation of BEAMS - Bayesian Estimation Applied to Multiple Species - showed how to use a dataset contaminated by points of multiple underlying types to perform unbiased parameter estimation. An example is cosmological parameter estimation from a photometric supernova sample contaminated by unknown Type Ibc and II supernovae. Where other methods require data cuts to increase purity, BEAMS uses all of the data points in conjunction with their probabilities of being each type. Here we extend the BEAMS formalism to allow for correlations between the data and the type probabilities of the objects as can occur in realistic cases. We show with simple simulations that this extension can be crucial, providing a 50% reduction in parameter estimation variance when such correlations do exist. We then go on to perform tests to quantify the importance of the type probabilities, one of which illustrates the effect of biasing the probabilities in various ways. Finally, a general presentation of the selection bias problem is given, and discussed in the context of future photometric supernova surveys and BEAMS, which lead to specific recommendations for future supernova surveys., 10 figures, 14 pages

We analyze the three-year SDSS-II Superernova (SN) Survey data and identify a sample of 1070 photometric SN Ia candidates based on their multi-band light curve data. This sample consists of SN candidates with no spectroscopic confirmation, with a subset of 210 candidates having spectroscopic redshifts of their host galaxies measured, while the remaining 860 candidates are purely photometric in their identification. We describe a method for estimating the efficiency and purity of photometric SN Ia classification when spectroscopic confirmation of only a limited sample is available, and demonstrate that SN Ia candidates from SDSS-II can be identified photometrically with ~91% efficiency and with a contamination of ~6%. Although this is the largest uniform sample of SN candidates to date for studying photometric identification, we find that a larger spectroscopic sample of contaminating sources is required to obtain a better characterization of the background events. A Hubble diagram using SN candidates with no spectroscopic confirmation, but with host galaxy spectroscopic redshifts, yields a distance modulus dispersion that is only ~20 - 40% larger than that of the spectroscopically-confirmed SN Ia sample alone with no significant bias. A Hubble diagram with purely photometric classification and redshift-distance measurements, however, exhibit biases that require further investigation for precision cosmology., The Astrophysical Journal (in press)

Future photometric supernova surveys will produce vastly more candidates than can be followed up spectroscopically, highlighting the need for effective classification methods based on lightcurves alone. Here we introduce boosting and kernel density estimation techniques which have minimal astrophysical input, and compare their performance on 20,000 simulated Dark Energy Survey lightcurves. We demonstrate that these methods are comparable to the best template fitting methods currently used, and in particular do not require the redshift of the host galaxy or candidate. However both methods require a training sample that is representative of the full population, so typical spectroscopic supernova subsamples will lead to poor performance. To enable the full potential of such blind methods, we recommend that representative training samples should be used and so specific attention should be given to their creation in the design phase of future photometric surveys., Comment: 19 pages, 41 figures. No changes. Additional material and summary video available at http://cosmoaims.wordpress.com/2010/09/30/boosting-for-supernova-classification/

We show that in cases of marginal detections (~ 3\sigma), such as that of Baryonic Acoustic Oscillations (BAO) in cosmology, the often-used Gaussian approximation to the full likelihood is very poor, especially beyond ~3\sigma. This can radically alter confidence intervals on parameters and implies that one cannot naively extrapolate 1\sigma-errorbars to 3\sigma, and beyond. We propose a simple fitting formula which corrects for this effect in posterior probabilities arising from marginal detections. Alternatively the full likelihood should be used for parameter estimation rather than the Gaussian approximation of a just mean and an error., Comment: 13 pages, 7 figures, includes extended discussion of the implications of Non-Gaussianity from marginal detections, along with explicit examples and an improved fitting formula to approximate the Non-Gaussian likelihood

We present a measurement of the volumetric Type Ia supernova (SN Ia) rate based on data from the Sloan Digital Sky Survey II (SDSS-II) Supernova Survey. The adopted sample of supernovae (SNe) includes 516 SNe Ia at redshift z \lesssim 0.3, of which 270 (52%) are spectroscopically identified as SNe Ia. The remaining 246 SNe Ia were identified through their light curves; 113 of these objects have spectroscopic redshifts from spectra of their host galaxy, and 133 have photometric redshifts estimated from the SN light curves. Based on consideration of 87 spectroscopically confirmed non-Ia SNe discovered by the SDSS-II SN Survey, we estimate that 2.04+1.61-0.95 % of the photometric SNe Ia may be misidentified. The sample of SNe Ia used in this measurement represents an order of magnitude increase in the statistics for SN Ia rate measurements in the redshift range covered by the SDSS-II Supernova Survey. If we assume a SN Ia rate that is constant at low redshift (z < 0.15), then the SN observations can be used to infer a value of the SN rate of rV = (2.69+0.34+0.21-0.30-0.01) x10^{-5} SNe yr^{-1} Mpc-3 (H0 /(70 km s^{-1} Mpc^{-1}))^{3} at a mean redshift of ~ 0.12, based on 79 SNe Ia of which 72 are spectroscopically confirmed. However, the large sample of SNe Ia included in this study allows us to place constraints on the redshift dependence of the SN Ia rate based on the SDSS-II Supernova Survey data alone. Fitting a power-law model of the SN rate evolution, r_V(z) = A_p x ((1 + z)/(1 + z0))^��, over the redshift range 0.0 < z < 0.3 with z0 = 0.21, results in A_p = (3.43+0.15-0.15) x 10^{-5} SNe yr^{-1} Mpc-3 (H0 /(70 km s^{-1} Mpc^{-1}))^{3} and ��= 2.04+0.90-0.89., 80 pages. v2 updates a few references.

This is the Users' Manual for the Fisher Matrix software Fisher4Cast and covers installation, GUI help, command line basics, code flow and data structure, as well as cosmological applications and extensions. Finally we discuss the extensive tests performed on the software., Comment: 37 pages, 27 figures. Software available at http://www.cosmology.org.za. Cosmological applications described in the accompanying release paper

This paper describes the Seventh Data Release of the Sloan Digital Sky Survey (SDSS), marking the completion of the original goals of the SDSS and the end of the phase known as SDSS-II. It includes 11, 663 deg2 of imaging data, with most of the ~2000 deg2 increment over the previous data release lying in regions of low Galactic latitude. The catalog contains five-band photometry for 357 million distinct objects. The survey also includes repeat photometry on a 120° long, 2fdg5 wide stripe along the celestial equator in the Southern Galactic Cap, with some regions covered by as many as 90 individual imaging runs. We include a co-addition of the best of these data, going roughly 2 mag fainter than the main survey over 250 deg2. The survey has completed spectroscopy over 9380 deg2 the spectroscopy is now complete over a large contiguous area of the Northern Galactic Cap, closing the gap that was present in previous data releases. There are over 1.6 million spectra in total, including 930, 000 galaxies, 120, 000 quasars, and 460, 000 stars. The data release includes improved stellar photometry at low Galactic latitude. The astrometry has all been recalibrated with the second version of the USNO CCD Astrograph Catalog, reducing the rms statistical errors at the bright end to 45 milliarcseconds per coordinate. We further quantify a systematic error in bright galaxy photometry due to poor sky determination ; this problem is less severe than previously reported for the majority of galaxies. Finally, we describe a series of improvements to the spectroscopic reductions, including better flat fielding and improved wavelength calibration at the blue end, better processing of objects with extremely strong narrow emission lines, and an improved determination of stellar metallicities.

Baryon Acoustic Oscillations (BAO) are frozen relics left over from the pre-decoupling universe. They are the standard rulers of choice for 21st century cosmology, providing distance estimates that are, for the first time, firmly rooted in well-understood, linear physics. This review synthesises current understanding regarding all aspects of BAO cosmology, from the theoretical and statistical to the observational, and includes a map of the future landscape of BAO surveys, both spectroscopic and photometric., Comment: 42 Pages, 21 colour figures. Extended version of chapter contributed to the book "Dark Energy", Ed. P. Ruiz-Lapuente, Cambridge University Press. Comments and suggested references welcomed. Supplementary slides available at http://cosmology.org.za/BAO_review_slides.ppt

We propose a "thought technique" for detecting Gravitational Waves using Einstein-Podolski-Rosen photon Entangled States. GWs decohere the entangled photon pairs, introduce a relative rotation and de-synchronize Alice and Bob's reference frames thus reducing the measured non-locality of correlated quanta described by Bell's inequalities. Gravitational Waves, distorting quantum encryption key statistics away from a pure white noise, act then as shadow eavesdroppers. The deviation from the intrinsic white-noise randomness of a Quantum Key Distribution process can reveal the presence of a gravitational wave by analyzing the emerging color distortions in the key. Photon entangled states provide the key advantage of revealing the polarization rotation introduced by GWs without the need of previously fixed reference frames, 5 pages, 1 figure

This paper describes the Sixth Data Release of the Sloan Digital Sky Survey. With this data release, the imaging of the northern Galactic cap is now complete. The survey contains images and parameters of roughly 287 million objects over 9583 deg2 , including scans over a large range of Galactic latitudes and longitudes. The survey also includes 1.27 million spectra of stars, galaxies, quasars, and blank sky (for sky subtraction) selected over 7425 deg2 . This release includes much more stellar spectroscopy than was available in previous data releases and also includes detailed estimates of stellar temperatures, gravities, and metallicities. The results of improved photometric calibration are now available, with uncertainties of roughly 1% in g, r, i, and z, and 2% in u, substantially better than the uncertainties in previous data releases. The spectra in this data release have improved wavelength and flux calibration, especially in the extreme blue and extreme red, leading to the qualitatively better determination of stellar types and radial velocities. The spectrophotometric fluxes are now tied to point-spread function magnitudes of stars rather than fiber magnitudes. This gives more robust results in the presence of seeing variations, but also implies a change in the spectrophotometric scale, which is now brighter by roughly 0.35 mag. Systematic errors in the velocity dispersions of galaxies have been fixed, and the results of two independent codes for determining spectral classifications and redshifts are made available. Additional spectral outputs are made available, including calibrated spectra from individual 15 minute exposures and the sky spectrum subtracted from each exposure. We also quantify a recently recognized underestimation of the brightnesses of galaxies of large angular extent due to poor sky subtraction; the bias can exceed 0.2 mag for galaxies brighter than r ¼ 14 mag.

This paper describes the Sixth Data Release of the Sloan Digital Sky Survey. With this data release, the imaging of the northern Galactic cap is now complete. The survey contains images and parameters of roughly 287 million objects over 9583 deg², including scans over a large range of Galactic latitudes and longitudes. The survey also includes 1.27 million spectra of stars, galaxies, quasars, and blank sky (for sky subtraction) selected over 7425 deg². This release includes much more stellar spectroscopy than was available in previous data releases and also includes detailed estimates of stellar temperatures, gravities, and metallicities. The results of improved photometric calibration are now available, with uncertainties of roughly 1% in g, r, i, and z, and 2% in u, substantially better than the uncertainties in previous data releases. The spectra in this data release have improved wavelength and flux calibration, especially in the extreme blue and extreme red, leading to the qualitatively better determination of stellar types and radial velocities. The spectrophotometric fluxes are now tied to point-spread function magnitudes of stars rather than fiber magnitudes. This gives more robust results in the presence of seeing variations, but also implies a change in the spectrophotometric scale, which is now brighter by roughly 0.35 mag. Systematic errors in the velocity dispersions of galaxies have been fixed, and the results of two independent codes for determining spectral classifications and redshifts are made available. Additional spectral outputs are made available, including calibrated spectra from individual 15 minute exposures and the sky spectrum subtracted from each exposure. We also quantify a recently recognized underestimation of the brightnesses of galaxies of large angular extent due to poor sky subtraction; the bias can exceed 0.2 mag for galaxies brighter than r = 14 mag.

This paper presents spectroscopy of supernovae discovered in the first season of the Sloan Digital Sky Survey-II Supernova Survey. This program searches for and measures multi-band light curves of supernovae in the redshift range z = 0.05 - 0.4, complementing existing surveys at lower and higher redshifts. Our goal is to better characterize the supernova population, with a particular focus on SNe Ia, improving their utility as cosmological distance indicators and as probes of dark energy. Our supernova spectroscopy program features rapid-response observations using telescopes of a range of apertures, and provides confirmation of the supernova and host-galaxy types as well as precise redshifts. We describe here the target identification and prioritization, data reduction, redshift measurement, and classification of 129 SNe Ia, 16 spectroscopically probable SNe Ia, 7 SNe Ib/c, and 11 SNe II from the first season. We also describe our efforts to measure and remove the substantial host galaxy contamination existing in the majority of our SN spectra., Comment: Accepted for publication in The Astronomical Journal(47pages, 9 figures)

Observed data is often contaminated by undiscovered interlopers, leading to biased parameter estimation. Here we present BEAMS (Bayesian Estimation Applied to Multiple Species) which significantly improves on the standard maximum likelihood approach in the case where the probability for each data point being `pure' is known. We discuss the application of BEAMS to future Type Ia supernovae (SNIa) surveys, such as LSST, which are projected to deliver over a million supernovae lightcurves without spectra. The multi-band lightcurves for each candidate will provide a probability of being Ia (pure) but the full sample will be significantly contaminated with other types of supernovae and transients. Given a sample of N supernovae with mean probability, P, of being Ia, BEAMS delivers parameter constraints equal to NP spectroscopically-confirmed SNIa. In addition BEAMS can be simultaneously used to tease apart different families of data and to recover properties of the underlying distributions of those families (e.g. the Type Ibc and II distributions). Hence BEAMS provides a unified classification and parameter estimation methodology which may be useful in a diverse range of problems such as photometric redshift estimation or, indeed, any parameter estimation problem where contamination is an issue., 12 pages, 5 figures. Minor revisions to match published version

We discuss how the baryon acoustic oscillation (BAO) signatures in the galaxy power spectrum can distinguish between modified gravity and the cosmological constant as the source of cosmic acceleration. To this end we consider a model characterized by a parameter n, which corresponds to the Dvali-Gabadadze-Porrati (DGP) model if n=2 and reduces to the standard spatially flat cosmological constant concordance model for n equal to infinity. We find that the different expansion histories of the modified gravity models systematically shifts the peak positions of BAO. A preliminary analysis using the current SDSS luminous red galaxy (LRG) sample indicates that the original DGP model is disfavored unless the matter density parameter exceeds 0.3. The constraints will be strongly tightened with future spectroscopic samples of galaxies at high redshifts. We demonstrate that WFMOS, in collaboration with other surveys such as Planck, will powerfully constrain modified gravity alternatives to dark energy as the explanation of cosmic acceleration

No abstract prepared.

Vast sound waves traveling through the relativistic plasma during the first million years of the universe imprint a preferred scale in the density of matter. We now have the ability to detect this characteristic fingerprint in the clustering of galaxies at various redshifts and use it to measure the acceleration of the expansion of the Universe. The Wide-Field Multi-Object Spectrograph (WFMOS) would use this test to shed significant light on the true nature of dark energy, the mysterious source of this cosmic acceleration. WFMOS would also revolutionise studies of the kinematics of the Milky Way and provide deep insights into the clustering of galaxies at redshifts up to z~4. In this article we discuss the recent progress in large galaxy redshift surveys and detail how WFMOS will help unravel the mystery of dark energy., Comment: 6 pages, pure pdf. An introduction to WFMOS and Baryon Acoustic Oscillations for a general audience

There are two basic ways to measure physical distances in cosmology: One based on standard candles and one based on standard rulers. Comparing current data for each method allows us to rule out axion-photon mixing and dust-extinction as the sources of supernova dimming and generally protects the case for cosmic acceleration from attacks based on loss of photons. The combined data constrains the energy densities in a LCDM model to 0.19 < Omega_m < 0.32 and 0.47 < Omega_Lambda < 0.82 (at 2\sigma) without recourse to any further data sets. Future data will improve on these limits and allow us to place constraints on more exotic physics., Comment: 4 pages, 1 figure, uses moriond.sty. To be published in the Proceedings of the XXXIX Rencontres de Moriond 'Exploring the Universe'

Optimising the major next-generation cosmological surveys (such as {\em SNAP, KAOS etc...}) is a key problem given our ignorance of the physics underlying cosmic acceleration and the plethora of surveys planned. We propose a Bayesian design framework which (1) maximises the discrimination power of a survey without assuming any underlying dark energy model, (2) finds the best niche survey geometry given current data and future competing experiments, (3) maximises the cross-section for serendipitous discoveries and (4) can be adapted to answer specific questions (such as `is dark energy dynamical?'). Integrated Parameter Space Optimisation (IPSO) is a design framework that integrates projected parameter errors over an entire dark energy parameter space and then extremises a figure of merit (such as Shannon entropy gain which we show is stable to off-diagonal covariance matrix perturbations) as a function of survey parameters using analytical, grid or MCMC techniques. We discuss examples where the optimisation can be performed analytically. IPSO is thus a general, model-independent and scalable framework that allows us to appropriately use prior information to design the best possible surveys., Comment: 15 pages, 6 colour figures. Significantly more discussion and illustrative examples

There are two basic ways to measure physical distances in cosmology: One based on standard candles and one based on standard rulers. Comparing current data for each method allows us to rule out axion-photon mixing and dust-extinction as the sources of supernova dimming and generally protects the case for cosmic acceleration from attacks based on loss of photons. The combined data constrains the energy densities in a LCDM model to 0.19 < Omega_m < 0.32 and 0.47 < Omega_Lambda < 0.82 (at 2��) without recourse to any further data sets. Future data will improve on these limits and allow us to place constraints on more exotic physics., 4 pages, 1 figure, uses moriond.sty. To be published in the Proceedings of the XXXIX Rencontres de Moriond 'Exploring the Universe'

Axion-photon mixing has been proposed as an alternative to acceleration as the explanation for supernovae dimming. We point out that the loss of photons due to this mixing will induce a strong asymmetry between the luminosity, d_L(z), and angular-diameter distance, d_A(z), since the latter is unaffected by mixing. In a first search for such an asymmetry we introduce a dimensionless mixing amplitude lambda so that lambda=0 if no photons are lost and lambda = 1 if axion-photon mixing occurs. The best-fit to SNIa and radio galaxy data is lambda = -0.3^{+0.6}_{-0.4} (95% CL), corresponding to an unphysical, negative, mixing length. This same argument limits the attenuation of light from supernovae due to dust. We show that future d_L and d_A data from SNAP and galaxy surveys such as DEEP2 and KAOS will detect or rule out mixing at more than 5-sigma, almost independently of the dark energy dynamics. Finally we discuss the constraints from the near maximal polarisation of the gamma-ray burst (GRB) GRB021206. Since mixing reduces the polarisation of distant sources, future observations of high redshift GRBs may provide orthogonal constraints on axion-photon mixing and related scenarios., 4 pages, 4 colour figures. Matches version to appear in ApJ. Significantly stronger constraints arising from new data: universe still accelerating at 3 sigma

The paper has been deeply reviewed and compeltely re-written., Comment: This paper has been withdrawn by the author and re-written by the autors after the interpretation of the new numerical results.

The dynamics of preheating after inflation has not been clearly understood yet.In particular, the issue of the generation of metric perturbations during preheating on super-horizon scale is still unsettled. Large scale perturbations may leave an imprint on the cosmic microwave background, or may become seeds for generation of primordial black holes. Hence, in order to make a connection between the particle physics models and cosmological observations, understanding the evolution of super-Hubble scale perturbations during preheating is important. Here, we propose an alternative treatment to handle this issue based on the so-called separate universe approach, which suggests less efficient amplification of super-Hubble modes during preheating than was expected before. We also point out an important issue which may have been overlooked in previous treatments., Comment: Proceedings of 12th JGRG meeting (Tokyo, Nov.25-28(2002))

Until now all well-studied dark energy models proposed have very large Compton wavelengths, $\lambda_c > 100 Mpc$ at all time and hence show no clustering on small scales. Evidence against this would disfavour both $\Lambda$CDM and quintessence. We show however, that large $\lambda_c$ is actually favoured by current CMB, LSS and SN1a data by comparing quintessence with a condensate scenario with the same background dynamics. The condensate scenario has a very small $\lambda_c$ before making a transition to negative pressure and large $\lambda_c$ and therefore exhibits a rich CMB phenomenology which roughly interpolates between $\Lambda$CDM and flat CDM models., Comment: 5 pages, 3 colour eps figures. Related to a talk given by BB at the XVIIIth IAP Colloquium "On the nature of Dark Energy", Paris, July 2002

We propose a new technique for detecting gravitational waves using Quantum Entangled STate (QUEST) technology. Gravitational waves reduce the non-locality of correlated quanta controlled by Bell's inequalities, distorting quantum encryption key statistics away from a pure white noise. Gravitational waves therefore act as shadow eavesdroppers. The resulting colour distortions can, at least in principle, be separated from noise and can differentiate both deterministic and stochastic sources., Comment: 10 pages, Revtex, 2 colour JPEG figures, high resolution versions of the figures and additional material available at http://sun1.sms.port.ac.uk/cosmos/users/bruce/mousetrap/

We demonstrate the parametric amplification of the stochastic gravitational wave background during inflationary reheating and during axion/moduli oscillations. This amplification enhances the detectability of the string/inflationary gravity wave signal, leaving a finger-print on the spectrum which might be found with future gravitational wave detectors., 5 pages, 1 eps figure, World Sci style file included. Talk at the 2nd Amaldi Conference on Gravitational Waves, CERN (1997). Eds. E. Coccia et al (World Scientific, Singapore)

The gravitational wave equations form a parametric resonance system during oscillatory reheating after inflation, when cast in terms of the electric and magnetic parts of the Weyl tensor. This is in direct analogy with preheating. For chaotic inflation with a quadratic potential, this analogy is exact. The resulting amplification of the gravitational wave spectrum begins in the broad resonance regime for chaotic inflation but quickly moves towards narrow resonance. This increases the amplitude and breaks the scale invariance of the tensor spectrum generated during inflation, an effect which may be detectable. The parametric amplification is absent, however, in 1st-order phase transition models and "warm" inflationary models with continuous entropy production., Comment: 10 pages, Revtex, 3 ps figures. Matched to published version

Includes bibliographical references., The main work of this thesis can be summarised as: ■ An implementation of canonical quantisation to the covariant and gauge-invariant approach to cosmological perturbations. Standard results are reproduced. We discuss the advantages of this formalism over non-covariant and non gauge-invariant formalisms. ■ A characterisation of linear gravitational waves in a covariant way is achieved. The evolution equations for the electric and magnetic parts of the Weyl tensor are shown to be of different order. In particular, the electric part appears to have a third order evolution equation, while the magnetic part has a second order evolution equation. It is shown that the "silent" nature of the evolution equations for irrotational dust can be extended to the case of vortical dust. This may be relevant for the endpoints of gravitational collapse since the vorticity begins to grow as soon as density contrast becomes non-linear, as is the case in galaxies, showing that the irrotational silent universes are unstable. The main problem in accepting such vortical silent universes lies in proving integrability of the equations which has not been achieved so far, even in the irrotational case. ■ A review of issues in the Cosmic Microwave Background (CMB) is given, focussing particularly on points such as ergodicity, decaying modes, foreground contamination, recombination, spectral distortions and polarisation of the CMB. ■ A review of methods in gravitational lensing is presented, together with a hierarchy of distance measures in cosmology, forming an introduction to the following two chapters. ■ A common belief that photon conservation implies that the all-sky averaged area distance in inhomogeneous universes must be that of the background, matter-averaged Robertson-Walker area distance is dis proven. This means that there will in general be gravitational lensing effects even on large angular scales. ■ The realistic situation in which gravitational lensing leads to caustic formation is discussed. It is claimed that this invalidates many accepted beliefs concerning high-redshift observations in inhomogeneous universes. One application of importance is the CMB. Possible implications are discussed. ■ Random Gaussian fields are ubiquitous in modern statistical physics, and particularly important in CMB studies. Here we give accurate analytical functions approximating ∫e⁻ˣ²dx, the simplest of which is just the kink soliton.

We develop and apply a fully covariant 1+3 electromagnetic analogy for gravity. The free gravitational field is covariantly characterized by the Weyl gravito-electric and gravito-magnetic spatial tensor fields, whose dynamical equations are the Bianchi identities. Using a covariant generalization of spatial vector algebra and calculus to spatial tensor fields, we exhibit the covariant analogy between the tensor Bianchi equations and the vector Maxwell equations. We identify gravitational source terms, couplings and potentials with and without electromagnetic analogues. The nonlinear vacuum Bianchi equations are shown to be invariant under covariant spatial duality rotation of the gravito-electric and gravito-magnetic tensor fields. We construct the super-energy density and super-Poynting vector of the gravitational field as natural U(1) group invariants, and derive their super-energy conservation equation. A covariant approach to gravito-electric/magnetic monopoles is also presented., 14 pages. Version to appear in Class. Quant. Grav

Since reinforcement learning algorithms have to fully solve a task in order to evaluate a set of hyperparameter values, conventional hyperparameter tuning methods can be highly sample inefficient and computationally expensive. Many widely used reinforcement learning architectures originate from scientific papers which include optimal hyperparameter values in the publications themselves, but do not indicate how the hyperparameter values were found. To address the issues related to hyperparameter tuning, three different experiments were investigated. In the first two experiments, Bayesian Optimisation and random search are compared. In the third and final experiment, the hyperparameter values found in second experiment are used to solve a more difficult reinforcement learning task, effectively performing hyperparameter transfer learning (later referred to as meta-transfer learning). The results from experiment 1 showed that there are certain scenarios in which Bayesian Optimisation outperforms random search for hyperparameter tuning, while the results of experiment 2 show that as more hyperparameters are simultaneously tuned, Bayesian Optimisation consistently finds better hyperparameter values than random search. However, BO took more than twice the amount of time to find these hyperparameter values than random search. Results from the third and final experiment indicate that hyperparameter values learned while tuning hyperparameters for a relatively easy to solve reinforcement learning task (Task A), can be used to solve a more complex task (Task B). With the available computing power for this thesis, hyperparameter optimisation was possible on the tasks in experiment 1 and experiment 2. This was not possible on the task in experiment 3, due to limited computing resources and the increased complexity of the reinforcement learning task in experiment 3, making the transfer of hyperparameters from one task (Task A) to the more difficult task (Task B) highly beneficial for solving the more computationally expensive task. The purpose of this work is to explore the effectiveness of Bayesian Optimisation as a hyperparameter tuning algorithm on the reinforcement learning algorithm NEAT's hyperparemters. An additional goal of this work is the experimental use of hyperparameter value transfer between reinforcement learning tasks, referred to in this work as Meta-Transfer Learning. This is introduced and discussed in greater detail in the Introduction chapter. All code used for this work is available in the repository: • https://github.com/maaxnaax/MSc_code

Anomaly detection is the process of finding unusual events in a given dataset. Anomaly detection is often performed on datasets with a fixed set of predefined features. As a result of this, if the normal features bear a close resemblance to the anomalous features, most anomaly detection algorithms exhibit poor performance. This work seeks to answer the question, can we deform these features so as to make the anomalies standout and hence improve the anomaly detection outcome? We employ a Deep Learning and an Active Learning framework to learn features for anomaly detection. In Active Learning, an Oracle (usually a domain expert) labels a small amount of data over a series of training rounds. The deep neural network is trained after each round to incorporate the feedback from the Oracle into the model. Results on the MNIST, CIFAR-10 and Galaxy Zoo datasets show that our algorithm, Ahunt, significantly outperforms other anomaly detection algorithms used on a fixed, static, set of features. Ahunt can therefore overcome a poor choice of features that happen to be suboptimal for detecting anomalies in the data, learning more appropriate features. We also explore the role of the loss function and Active Learning query strategy, showing these are important, especially when there is a significant variation in the anomalies.

Radio Frequency Interference (RFI) is a massive problem for radio observatories around the world. Due to the growth of telecommunications and air travel RFI is increasing exactly when the world's radio telescopes are increasing significantly in sensitivity, making RFI one of the most pressing problems for astronomy in the era of the Square Kilometre Array (SKA). Traditionally RFI is dealt with through simple algorithms that remove unexpected rapid changes but the recent explosion of machine learning and artificial intelligence (AI) provides an exciting opportunity for pushing the state-of-the-art in RFI excision. Unfortunately, due to the lack of training data for which the true RFI contamination is known, it is impossible to reliably train and compare machine learning algorithms for RFI excision on radio telescope arrays currently. To address this stumbling block we present RFIsim, a radio interferometry simulator that includes the telescope properties of the MeerKAT array, a sky model based on previous radio surveys coupled with an RFI model designed to reproduce actual RFI seen at the MeerKAT site. We perform an indepth comparison of the simulator results with real observations using the MeerKAT telescope and show that RFIsim produces visibilities that mimic those produced by real observations very well. Finally, we describe how the data was key in the development of a new state-of-the-art deep learning RFI flagging algorithm in Vafaei et al. (2020.) [69] In particular, this work demonstrates that transfer learning from simulation to real data is an effective way to leverage the power of machine learning for RFI flagging in real-world observatories.

The primary objective of this thesis is to develop rigorous Bayesian tools for common statistical challenges arising in modern science where there is a heightened demand for precise inference in the presence of large, known uncertainties. This thesis explores in detail two arenas where this manifests. The first is the development and testing of a unified Bayesian anomaly detection and classification framework (BADAC) which allows principled anomaly detection in the presence of measurement uncertainties, which are rarely incorporated into machine learning algorithms. BADAC deals with uncertainties by marginalising over the unknown, true value of the data. Using simulated data with Gaussian noise as an example, BADAC is shown to be superior to standard algorithms in both classification and anomaly detection performance in the presence of uncertainties. Additionally, BADAC provides well-calibrated classification probabilities, valuable for use in scientific pipelines. BADAC is therefore ideal where computational cost is not a limiting factor and statistical rigour is important. We discuss approximations to speed up BADAC, such as the use of Gaussian processes, and finally introduce a new metric, the Rank-Weighted Score (RWS), that is particularly suited to evaluating an algorithm's ability to detect anomalies. The second major exploration in this thesis presents methods for rigorous statistical inference in the presence of classification uncertainties and errors. Although this is explored specifically through supernova cosmology, the context is general. Supernova cosmology without spectra will be an important component of future surveys due to massive increases in data volumes in next-generation surveys such as from the Vera C. Rubin Observatory. This lack of supernova spectra results both in uncertainty in the redshifts and type of the supernova, which if ignored, leads to significantly biased estimates of cosmological parameters. We present a hierarchical Bayesian formalism, zBEAMS, which addresses this problem by marginalising over the unknown or uncertain supernova redshifts and types to produce unbiased cosmological estimates that are competitive with supernova data with fully spectroscopically confirmed redshifts. zBEAMS thus provides a unified treatment of both photometric redshifts, classification uncertainty and host galaxy misidentification, effectively correcting the inevitable contamination in the Hubble diagram with little or no loss of statistical power.

Radio signals from astronomical sources are extremely weak and easily distorted/- corrupted or overwhelmed by man-made radio signals such as cellphones, satellites, aircraft and telescope electronics. These Radio Frequency Interference (RFI) are increasingly threatening radio observatories due to our increasingly technological world. To detect and mitigate RFI, observatories need to understand their RFI environment, what contributes to it and how it is changing. While there are few dedicated RFI monitoring systems on the MeerKAT site, the most sensitive RFI detector is the MeerKAT array itself. In this thesis we use approximately 1500 hours of MeerKAT observations to create a multi-dimensional view of the RFI at the MeerKAT site. Here we investigate a probabilistic approach to characterise the RFI environment around the MeerKAT radio telescope. In order to achieve our goal, we propose the MeerKAT Historical Probability of RFI (KATHPRFI) framework. We produced the high level requirements of the KATHPRFI framework driven by the needs of the MeerKAT users. The design approach and the design decision of the framework is presented that cover both the software and hardware constraints. The KATHPRFI produces a 5-dimensional array of the RFI probability as measured by the MeerKAT telescope during the commissioning phase (May 2018 - December 2018) for each observation file. From the 5-D array, we extracted various statistics and characterised the RFI environment around MeerKAT site. We found that there is a correlation between RFI occupancy and the time of the day which is most probably related to human activities. Furthermore, we found a correlation between the time of the day and flights passing over a region of site. Our results showed that the highest probability of RFI points towards a region including nearby towns. The results obtained are consistent with the argument that the major RFI sources for MeerKAT site are the Global Positioning System (GPS) satellite, flight Distance Measurement Equipment (DME) and the Global System for Mobile Communications (GSM). Our data also showed that the RFI occupancy decreases with an increase of baseline length, this is a result of moving RFI sources with respect to the static sky. Therefore, the phase of the RFI changes rapidly on long baselines compared to short baselines. As a result when a correlation is carried out the RFI amplitude will vanish less on short baselines compared to the long baselines. Our results provide the first highly detailed view of the MeerKAT RFI environment allowing us to track the historical evolution of the RFI, both on average, and as a function of frequency, baseline and direction. With historical baselines known, one can also provide alerts about sudden changes. This could be due to new sources of RFI or stem from any outliers in the data, which could signal telescope or correlator issues. Hence the KATHPRFI framework also provides a window into the operational health of the telescope.