Search

As deep neural networks (DNNs) become increasingly common, concerns about their robustness do as well. A longstanding problem for deployed DNNs is their behavior in the face of unfamiliar inputs; specifically, these models tend to be overconfident and incorrect when encountering out-of-distribution (OOD) examples. In this work, we present a topological approach to characterizing OOD examples using latent layer embeddings from DNNs. Our goal is to identify topological features, referred to as landmarks, that indicate OOD examples. We conduct extensive experiments on benchmark datasets and a realistic DNN model, revealing a key insight for OOD detection. Well-trained DNNs have been shown to induce a topological simplification on training data for simple models and datasets; we show that this property holds for realistic, large-scale test and training data, but does not hold for OOD examples. More specifically, we find that the average lifetime (or persistence) of OOD examples is statistically longer than that of training or test examples. This indicates that DNNs struggle to induce topological simplification on unfamiliar inputs. Our empirical results provide novel evidence of topological simplification in realistic DNNs and lay the groundwork for topologically-informed OOD detection strategies., Comment: 24 pages, 13 figures

We present the MULTIMODAL UNIVERSE, a large-scale multimodal dataset of scientific astronomical data, compiled specifically to facilitate machine learning research. Overall, the MULTIMODAL UNIVERSE contains hundreds of millions of astronomical observations, constituting 100\,TB of multi-channel and hyper-spectral images, spectra, multivariate time series, as well as a wide variety of associated scientific measurements and "metadata". In addition, we include a range of benchmark tasks representative of standard practices for machine learning methods in astrophysics. This massive dataset will enable the development of large multi-modal models specifically targeted towards scientific applications. All codes used to compile the MULTIMODAL UNIVERSE and a description of how to access the data is available at https://github.com/MultimodalUniverse/MultimodalUniverse, Comment: Accepted at NeurIPS Datasets and Benchmarks track

Machine learning has great potential to increase crop production and resilience to climate change. Accurate maps of where crops are grown are a key input to a number of downstream policy and research applications. In this proposal, we present preliminary work showing that it is possible to accurately classify crops from time series derived from Sentinel 1 and 2 satellite imagery in Mexico using a pixel-based binary crop/non-crop time series transformer model. We also find preliminary evidence that meta-learning approaches supplemented with data from similar agro-ecological zones may improve model performance. Due to these promising results, we propose further development of this method with the goal of accurate multi-class crop classification in Jalisco, Mexico via meta-learning with a dataset comprising similar agro-ecological zones.

We propose the unified BRAVO challenge to benchmark the reliability of semantic segmentation models under realistic perturbations and unknown out-of-distribution (OOD) scenarios. We define two categories of reliability: (1) semantic reliability, which reflects the model's accuracy and calibration when exposed to various perturbations; and (2) OOD reliability, which measures the model's ability to detect object classes that are unknown during training. The challenge attracted nearly 100 submissions from international teams representing notable research institutions. The results reveal interesting insights into the importance of large-scale pre-training and minimal architectural design in developing robust and reliable semantic segmentation models., Comment: ECCV 2024 proceeding paper of the BRAVO challenge 2024, see https://benchmarks.elsa-ai.eu/?ch=1&com=introduction Corrected numbers in Tables 1,3,4,5 and 10

The performance of fiber-fed astronomical spectrographs is highly influenced by the properties of fibers. The near-field and far-field scrambling characteristics have a profound impact on the line spread function (LSF) of the spectra. Focal ratio degradation (FRD) influences the output beam size, thereby affecting the throughput, as well as the size of the collimator and dispersion elements. While previous research has indicated that these properties depend on the shape of the fiber core and showed that non-circular core fibers can yield uniform near-field scrambling, the result remains inconclusive for far-field. In this study, we investigate the near-field and far-field scrambling properties, along with the FRD, of 50-micron core fibers with different core geometries. We find that in addition to excellent near-field scrambling, octagonal-core fibers can also produce more uniform far-field output when compared to circular-core fibers. They also have less FRD effect when being fed with a f/3 beam., Comment: 13 pages, 11 figures, SPIE proceedings, Ground-based and Airborne Instrumentation for Astronomy X

The exponential growth of astronomical literature poses significant challenges for researchers navigating and synthesizing general insights or even domain-specific knowledge. We present Pathfinder, a machine learning framework designed to enable literature review and knowledge discovery in astronomy, focusing on semantic searching with natural language instead of syntactic searches with keywords. Utilizing state-of-the-art large language models (LLMs) and a corpus of 350,000 peer-reviewed papers from the Astrophysics Data System (ADS), Pathfinder offers an innovative approach to scientific inquiry and literature exploration. Our framework couples advanced retrieval techniques with LLM-based synthesis to search astronomical literature by semantic context as a complement to currently existing methods that use keywords or citation graphs. It addresses complexities of jargon, named entities, and temporal aspects through time-based and citation-based weighting schemes. We demonstrate the tool's versatility through case studies, showcasing its application in various research scenarios. The system's performance is evaluated using custom benchmarks, including single-paper and multi-paper tasks. Beyond literature review, Pathfinder offers unique capabilities for reformatting answers in ways that are accessible to various audiences (e.g. in a different language or as simplified text), visualizing research landscapes, and tracking the impact of observatories and methodologies. This tool represents a significant advancement in applying AI to astronomical research, aiding researchers at all career stages in navigating modern astronomy literature., Comment: 25 pages, 9 figures, submitted to AAS jorunals. Comments are welcome, and the tools mentioned are available online at https://pfdr.app

This work presents AstroPT, an autoregressive pretrained transformer developed with astronomical use-cases in mind. The AstroPT models presented here have been pretrained on 8.6 million $512 \times 512$ pixel $grz$-band galaxy postage stamp observations from the DESI Legacy Survey DR8. We train a selection of foundation models of increasing size from 1 million to 2.1 billion parameters, and find that AstroPT follows a similar saturating log-log scaling law to textual models. We also find that the models' performances on downstream tasks as measured by linear probing improves with model size up to the model parameter saturation point. We believe that collaborative community development paves the best route towards realising an open source `Large Observation Model' -- a model trained on data taken from the observational sciences at the scale seen in natural language processing. To this end, we release the source code, weights, and dataset for AstroPT under the MIT license, and invite potential collaborators to join us in collectively building and researching these models., Comment: 12 pages, 4 figures, 1 table. Code available at https://github.com/Smith42/astroPT

We propose a novel distributional regression model for a multivariate response vector based on a copula process over the covariate space. It uses the implicit copula of a Gaussian multivariate regression, which we call a ``regression copula''. To allow for large covariate vectors their coefficients are regularized using a novel multivariate extension of the horseshoe prior. Bayesian inference and distributional predictions are evaluated using efficient variational inference methods, allowing application to large datasets. An advantage of the approach is that the marginal distributions of the response vector can be estimated separately and accurately, resulting in predictive distributions that are marginally-calibrated. Two substantive applications of the methodology highlight its efficacy in multivariate modeling. The first is the econometric modeling and prediction of half-hourly regional Australian electricity prices. Here, our approach produces more accurate distributional forecasts than leading benchmark methods. The second is the evaluation of multivariate posteriors in likelihood-free inference (LFI) of a model for tree species abundance data, extending a previous univariate regression copula LFI method. In both applications, we demonstrate that our new approach exhibits a desirable marginal calibration property.

We explore the potential of enhancing LLM performance in astronomy-focused question-answering through targeted, continual pre-training. By employing a compact 7B-parameter LLaMA-2 model and focusing exclusively on a curated set of astronomy corpora -- comprising abstracts, introductions, and conclusions -- we achieve notable improvements in specialized topic comprehension. While general LLMs like GPT-4 excel in broader question-answering scenarios due to superior reasoning capabilities, our findings suggest that continual pre-training with limited resources can still enhance model performance on specialized topics. Additionally, we present an extension of AstroLLaMA: the fine-tuning of the 7B LLaMA model on a domain-specific conversational dataset, culminating in the release of the chat-enabled AstroLLaMA for community use. Comprehensive quantitative benchmarking is currently in progress and will be detailed in an upcoming full paper. The model, AstroLLaMA-Chat, is now available at https://huggingface.co/universeTBD, providing the first open-source conversational AI tool tailored for the astronomy community., Comment: 4 pages, 1 figure, model is available at https://huggingface.co/universeTBD, published in RNAAS

Forests of the Earth are a vital carbon sink while providing an essential habitat for biodiversity. Vegetation productivity (VP) is a critical indicator of carbon uptake in the atmosphere. The leaf area index is a crucial vegetation index used in VP estimation. This work proposes to predict the leaf area index (LAI) using climate variables to better understand future productivity dynamics; our approach leverages the capacities of the V-Net architecture for spatiotemporal LAI prediction. Preliminary results are well-aligned with established quality standards of LAI products estimated from Earth observation data. We hope that this work serves as a robust foundation for subsequent research endeavours, particularly for the incorporation of prediction attribution methodologies, which hold promise for elucidating the underlying climate change drivers of global vegetation productivity., Comment: 7 pages, 2 figures, submitted to Tackling Climate Change with Machine Learning at NeurIPS 2023

Planets grow in rotating disks of dust and gas around forming stars, some of which can subsequently collide in giant impacts after the gas component is removed from the disk. Monitoring programs with the warm Spitzer mission have recorded significant and rapid changes in mid-infrared output for several stars, interpreted as variations in the surface area of warm dusty material ejected by planetary-scale collisions and heated by the central star: e.g., NGC 2354-ID8, HD 166191 and V844 Persei. Here we report combined observations of the young (about 300 Myr), solar-like star ASASSN-21qj: an infrared brightening consistent with a blackbody temperature of 1000 K and a luminosity of 4 percent of that of the star lasting for about 1000 days, partially overlapping in time with a complex and deep wavelength-dependent optical eclipse that lasted for about 500 days. The optical eclipse started 2.5 years after the infrared brightening, implying an orbital period of at least that duration. These observations are consistent with a collision between two exoplanets of several to tens of Earth masses at 2 to 16 au from the central star. Such an impact produces a hot, highly-extended post-impact remnant with sufficient luminosity to explain the infrared observations. Transit of the impact debris, sheared by orbital motion into a long cloud, causes the subsequent complex eclipse of the host star., Comment: 28 pages, 9 figures, 2 tables, author's Accepted Manuscript version, reproducible workflow built with showyourwork; open-source code can be found at https://github.com/mkenworthy/ASASSN-21qj-collision/

In copula models the marginal distributions and copula function are specified separately. We treat these as two modules in a modular Bayesian inference framework, and propose conducting modified Bayesian inference by "cutting feedback". Cutting feedback limits the influence of potentially misspecified modules in posterior inference. We consider two types of cuts. The first limits the influence of a misspecified copula on inference for the marginals, which is a Bayesian analogue of the popular Inference for Margins (IFM) estimator. The second limits the influence of misspecified marginals on inference for the copula parameters by using a pseudo likelihood of the ranks to define the cut model. We establish that if only one of the modules is misspecified, then the appropriate cut posterior gives accurate uncertainty quantification asymptotically for the parameters in the other module. Computation of the cut posteriors is difficult, and new variational inference methods to do so are proposed. The efficacy of the new methodology is demonstrated using both simulated data and a substantive multivariate time series application from macroeconomic forecasting. In the latter, cutting feedback from misspecified marginals to a 1096 dimension copula improves posterior inference and predictive accuracy greatly, compared to conventional Bayesian inference.

This article proposes a framework for content moderation based on a decentralized market where no one party, neither governments nor firms, controls the flow of information., Comment: 3 pages. This is published by CACM but the authors hold (c) and are happy to license under the listed Creative Commons CC:BY-SA license

We introduce EarthPT -- an Earth Observation (EO) pretrained transformer. EarthPT is a 700 million parameter decoding transformer foundation model trained in an autoregressive self-supervised manner and developed specifically with EO use-cases in mind. We demonstrate that EarthPT is an effective forecaster that can accurately predict future pixel-level surface reflectances across the 400-2300 nm range well into the future. For example, forecasts of the evolution of the Normalised Difference Vegetation Index (NDVI) have a typical error of approximately 0.05 (over a natural range of -1 -> 1) at the pixel level over a five month test set horizon, out-performing simple phase-folded models based on historical averaging. We also demonstrate that embeddings learnt by EarthPT hold semantically meaningful information and could be exploited for downstream tasks such as highly granular, dynamic land use classification. Excitingly, we note that the abundance of EO data provides us with -- in theory -- quadrillions of training tokens. Therefore, if we assume that EarthPT follows neural scaling laws akin to those derived for Large Language Models (LLMs), there is currently no data-imposed limit to scaling EarthPT and other similar `Large Observation Models.', Comment: 7 pages, 4 figures, accepted to NeurIPS CCAI workshop at https://www.climatechange.ai/papers/neurips2023/2 . Code available at https://github.com/aspiaspace/EarthPT

Skew-t copula models are attractive for the modeling of financial data because they allow for asymmetric and extreme tail dependence. We show that the copula implicit in the skew-t distribution of Azzalini and Capitanio (2003) allows for a higher level of pairwise asymmetric dependence than two popular alternative skew-t copulas. Estimation of this copula in high dimensions is challenging, and we propose a fast and accurate Bayesian variational inference (VI) approach to do so. The method uses a generative representation of the skew-t distribution to define an augmented posterior that can be approximated accurately. A stochastic gradient ascent algorithm is used to solve the variational optimization. The methodology is used to estimate skew-t factor copula models with up to 15 factors for intraday returns from 2017 to 2021 on 93 U.S. equities. The copula captures substantial heterogeneity in asymmetric dependence over equity pairs, in addition to the variability in pairwise correlations. In a moving window study we show that the asymmetric dependencies also vary over time, and that intraday predictive densities from the skew-t copula are more accurate than those from benchmark copula models. Portfolio selection strategies based on the estimated pairwise asymmetric dependencies improve performance relative to the index.

As deep learning-based computer vision algorithms continue to advance the state of the art, their robustness to real-world data continues to be an issue, making it difficult to bring an algorithm from the lab to the real world. Ensemble-based uncertainty estimation approaches such as Monte Carlo Dropout have been successfully used in many applications in an attempt to address this robustness issue. Unfortunately, it is not always clear if such ensemble-based approaches can be applied to a new problem domain. This is the case with panoptic segmentation, where the structure of the problem and architectures designed to solve it means that unlike image classification or even semantic segmentation, the typical solution of using a mean across samples cannot be directly applied. In this paper, we demonstrate how ensemble-based uncertainty estimation approaches such as Monte Carlo Dropout can be used in the panoptic segmentation domain with no changes to an existing network, providing both improved performance and more importantly a better measure of uncertainty for predictions made by the network. Results are demonstrated quantitatively and qualitatively on the COCO, KITTI-STEP and VIPER datasets., Comment: 29 pages, 9 figures including Supplementary Material. Revision includes additional comparisons and clarifications

Weyl fermions possess the property of spin-momentum locking: the expectation value of the spin is parallel or antiparallel to the momentum at any given point in the Brillouin zone in the vicinity of a Weyl node. This is a direct consequence of the fact that Weyl nodes are monopoles of the Berry curvature, and in this sense an expression of the nontrivial Weyl electronic structure topology. Thanks to this property, an isolated Weyl node produces a large spin accumulation in response to a charge current, $\hbar/2$ per electron, similar to surface states of time-reversal invariant topological insulators. However, in bulk Weyl semimetals, the nodes must occur in pairs of opposite chirality and, when the nodes are at the same energy, the effect cancels out. Here we show that this cancellation is avoided in chiral semimetals, in which Weyl nodes of opposite chirality occur at different energies due to broken mirror symmetry. We find that the spin accumulation is maximized when the Fermi energy coincides with one of the nodes in a pair and reaches the same value as for an isolated node in this case. Moreover, we demonstrate the existence of a distinct magnetoresistance mechanism, closely related to this current-induced spin accumulation., Comment: 7 pages, 4 figures, published version

Contrary to Google Search's mission of delivering information from "many angles so you can form your own understanding of the world," we find that Google and its most prominent returned results - Wikipedia and YouTube - simply reflect a narrow set of culturally dominant views tied to the search language for complex topics like "Buddhism," "Liberalism," "colonization," "Iran" and "America." Simply stated, they present, to varying degrees, distinct information across the same search in different languages, a phenomenon we call language bias. This paper presents evidence and analysis of language bias and discusses its larger social implications. We find that our online searches and emerging tools like ChatGPT turn us into the proverbial blind person touching a small portion of an elephant, ignorant of the existence of other cultural perspectives. Language bias sets a strong yet invisible cultural barrier online, where each language group thinks they can see other groups through searches, but in fact, what they see is their own reflection.

Condensation and sublimation of ices at the surface of the planet is a key part of both the Martian H$_2$O and CO$_2$ cycles, either from a seasonal or diurnal aspect. While most of the ice is located within the polar caps, surface frost is known to be formed during nighttime down to equatorial latitudes. Here, we use data from the Emirates Mars Infrared Spectrometer (EMIRS) onboard the Emirates Mars Mission (EMM) to monitor the diurnal and seasonal evolution of the ices at the surface of Mars over almost one Martian year. The unique local time coverage provided by the instrument allows us to observe the apparition of equatorial CO$_2$ frost in the second half of the Martian night around the equinoxes, to its sublimation at sunrise.

Stochastic models with global parameters and latent variables are common, and for which variational inference (VI) is popular. However, existing methods are often either slow or inaccurate in high dimensions. We suggest a fast and accurate VI method for this case that employs a well-defined natural gradient variational optimization that targets the joint posterior of the global parameters and latent variables. It is a hybrid method, where at each step the global parameters are updated using the natural gradient and the latent variables are generated from their conditional posterior. A fast to compute expression for the Tikhonov damped Fisher information matrix is used, along with the re-parameterization trick, to provide a stable natural gradient. We apply the approach to deep mixed models, which are an emerging class of Bayesian neural networks with random output layer coefficients to allow for heterogeneity. A range of simulations show that using the natural gradient is substantially more efficient than using the ordinary gradient, and that the approach is faster and more accurate than two cutting-edge natural gradient VI methods. In a financial application we show that accounting for industry level heterogeneity using the deep mixed model improves the accuracy of asset pricing models. MATLAB code to implement the method can be found at: https://github.com/WeibenZhang07/NG-HVI.

The X-ray Multi-mirror Mission (XMM-Newton) provides simultaneous non-dispersive spectroscopic X-ray imaging and timing, medium resolution dispersive X-ray spectroscopy and optical/UV imaging, spectroscopy and timing. In combination, the imaging cameras offer an effective area over the energy range from 150 eV to 12 keV of up to 2500 cm$^2$ at 1.5 keV and $\sim$1800 cm$^2$ at 5 keV. The gratings cover an energy range from 0.4 keV to 2.2 keV with a combined effective area of up to 120 cm$^2$ at 0.8 keV. XMM-Newton offers unique opportunities for a wide variety of sensitive X-ray observations accompanied by simultaneous optical/UV measurements. The majority of XMM-Newton's observing time is made available to the astronomical community by peer-reviewed Announcements of Opportunity. The scientific exploitation of XMM-Newton data is aided by an observatory-class X-ray facility which provides analysis software, pipeline processing, calibration and catalogue generation. Around 380 refereed papers based on XMM-Newton data are published each year with a high fraction of papers reporting transformative scientific results., Comment: 37 pages, 28 figures, Invited chapter for {\it Handbook of X-ray and Gamma-ray Astrophysics} (Eds. C. Bambi and A. Santangelo, Springer Singapore, expected in 2022

In this review, we explore the historical development and future prospects of artificial intelligence (AI) and deep learning in astronomy. We trace the evolution of connectionism in astronomy through its three waves, from the early use of multilayer perceptrons, to the rise of convolutional and recurrent neural networks, and finally to the current era of unsupervised and generative deep learning methods. With the exponential growth of astronomical data, deep learning techniques offer an unprecedented opportunity to uncover valuable insights and tackle previously intractable problems. As we enter the anticipated fourth wave of astronomical connectionism, we argue for the adoption of GPT-like foundation models fine-tuned for astronomical applications. Such models could harness the wealth of high-quality, multimodal astronomical data to serve state-of-the-art downstream tasks. To keep pace with advancements driven by Big Tech, we propose a collaborative, open-source approach within the astronomy community to develop and maintain these foundation models, fostering a symbiotic relationship between AI and astronomy that capitalizes on the unique strengths of both fields., Comment: 75 pages, 327 references, 32 figures. Review accepted in Royal Society Open Science

Temperature profiles retrieved using the first set of data of the Emirates Mars InfraRed Spectrometer (EMIRS) obtained during the science phase of the Emirates Mars Mission (EMM) are used for the analysis of migrating thermal tides in the Martian atmosphere. The selected data cover a solar longitude (LS) range of 60{\deg}-90{\deg} of Martian Year (MY) 36. The novel orbit design of the Hope Probe leads to a good geographic and local time coverage that significantly improves the analysis. Wave mode decomposition suggests dominant diurnal tide and important semi-diurnal tide with maximal amplitudes of 6K and 2K, respectively, as well as the existence of ~0.5K ter-diurnal tide. The results agree well with predictions by the Mars Planetary Climate Model (PCM), but the observed diurnal tide has an earlier phase (3h), and the semi-diurnal tide has an unexpectedly large wavelength (~200km)., Comment: 11 pages, 5 figures, accepted by GRL special issue "The First Results from the Emirates Mars Mission (EMM)"

Two fibre integral field units (IFU) are being built in the SAAO fibre-lab for the Robert Stobie Spectrograph's visible arm and the future red arm. Each IFU sits in its own slit-mask cassette and is referred to as a slit-mask IFU (SMI). They will be available some time in 2022. The smaller, 200 micron fibre IFU has 309 X 0.9 arcsec diameter spatial elements covering an elongated hexagonal footprint of 18 X 23 arcsec. The larger, 400 micron fibre IFU has 178 X 1.8 arcsec diameter spatial elements covering an on-sky area of 21 X 44 arcsec. In both cases there are two groups of 13 fibres offset by roughly 50 arcsec on either side of the primary array to sample sky. The 1.8 and 0.9 arcsec spatial resolution SMIs provide median spectral resolution of 1200 and 2400 respectively at H alpha wavelengths in the low resolution mode covering 320 to 740 nm bandpass. At a higher grating angle the SMI will deliver spectral resolution up to 5000 and 10000 with 400 and 200 micron core fibre respectively. A future red-arm will extend the simultaneous wavelength coverage up to 900 nm at a median resolution of 3000/6000 for the same flavors of IFUs. SMIs are inserted in the same fashion as the existing longslit cassettes at the SALT focal plane. Prismatic fold mirrors direct the focal plane into the fibre IFU and then back into the RSS collimator after the fibres are routed 180 deg within the cassette and formatted into a pseudo-slit. Fold prisms ensure that the spectrograph collimator continues to see the same focal plane. In this paper we describe the design, fabrication, assembly and characterization of Slit Mask IFUs.