Search

BabyIAXO is the intermediate stage of the International Axion Observatory (IAXO) to be hosted at DESY. Its primary goal is the detection of solar axions following the axion helioscope technique. Axions are converted into photons in a large magnet that is pointing to the sun. The resulting X-rays are focused by appropriate X-ray optics and detected by sensitive low-background detectors placed at the focal spot. The aim of this article is to provide an accurate quantitative description of the different components (such as the magnet, optics, and X-ray detectors) involved in the detection of axions. Our efforts have focused on developing robust and integrated software tools to model these helioscope components, enabling future assessments of modifications or upgrades to any part of the IAXO axion helioscope and evaluating the potential impact on the experiment's sensitivity. In this manuscript, we demonstrate the application of these tools by presenting a precise signal calculation and response analysis of BabyIAXO's sensitivity to the axion-photon coupling. Though focusing on the Primakoff solar flux component, our virtual helioscope model can be used to test different production mechanisms, allowing for direct comparisons within a unified framework., Comment: 36 pages, 18 figures, 4 tables, Submitted to JHEP

Context: The emergence of Large Language Models (LLMs) has significantly transformed Software Engineering (SE) by providing innovative methods for analyzing software repositories. Objectives: Our objective is to establish a practical framework for future SE researchers needing to enhance the data collection and dataset while conducting software repository mining studies using LLMs. Method: This experience report shares insights from two previous repository mining studies, focusing on the methodologies used for creating, refining, and validating prompts that enhance the output of LLMs, particularly in the context of data collection in empirical studies. Results: Our research packages a framework, coined Prompt Refinement and Insights for Mining Empirical Software repositories (PRIMES), consisting of a checklist that can improve LLM usage performance, enhance output quality, and minimize errors through iterative processes and comparisons among different LLMs. We also emphasize the significance of reproducibility by implementing mechanisms for tracking model results. Conclusion: Our findings indicate that standardizing prompt engineering and using PRIMES can enhance the reliability and reproducibility of studies utilizing LLMs. Ultimately, this work calls for further research to address challenges like hallucinations, model biases, and cost-effectiveness in integrating LLMs into workflows.

Background: Open-Source Pre-Trained Models (PTMs) and datasets provide extensive resources for various Machine Learning (ML) tasks, yet these resources lack a classification tailored to Software Engineering (SE) needs. Aims: We apply an SE-oriented classification to PTMs and datasets on a popular open-source ML repository, Hugging Face (HF), and analyze the evolution of PTMs over time. Method: We conducted a repository mining study. We started with a systematically gathered database of PTMs and datasets from the HF API. Our selection was refined by analyzing model and dataset cards and metadata, such as tags, and confirming SE relevance using Gemini 1.5 Pro. All analyses are replicable, with a publicly accessible replication package. Results: The most common SE task among PTMs and datasets is code generation, with a primary focus on software development and limited attention to software management. Popular PTMs and datasets mainly target software development. Among ML tasks, text generation is the most common in SE PTMs and datasets. There has been a marked increase in PTMs for SE since 2023 Q2. Conclusions: This study underscores the need for broader task coverage to enhance the integration of ML within SE practices., Comment: 5 pages, 8 figures

Background: Given the fast-paced nature of today's technology, which has surpassed human performance in tasks like image classification, visual reasoning, and English understanding, assessing the impact of Machine Learning (ML) on energy consumption is crucial. Traditionally, ML projects have prioritized accuracy over energy, creating a gap in energy consumption during model inference. Aims: This study aims to (i) analyze image classification datasets and pre-trained models, (ii) improve inference efficiency by comparing optimized and non-optimized models, and (iii) assess the economic impact of the optimizations. Method: We conduct a controlled experiment to evaluate the impact of various PyTorch optimization techniques (dynamic quantization, torch.compile, local pruning, and global pruning) to 42 Hugging Face models for image classification. The metrics examined include GPU utilization, power and energy consumption, accuracy, time, computational complexity, and economic costs. The models are repeatedly evaluated to quantify the effects of these software engineering tactics. Results: Dynamic quantization demonstrates significant reductions in inference time and energy consumption, making it highly suitable for large-scale systems. Additionally, torch.compile balances accuracy and energy. In contrast, local pruning shows no positive impact on performance, and global pruning's longer optimization times significantly impact costs. Conclusions: This study highlights the role of software engineering tactics in achieving greener ML models, offering guidelines for practitioners to make informed decisions on optimization methods that align with sustainability goals.

We present a catalog of stellar parameters (effective temperature $T_{\rm eff}$, surface gravity $\log g$, age, and metallicity [Fe/H]) and elemental-abundance ratios ([C/Fe], [Mg/Fe], and [$\alpha$/Fe]) for some five million stars (4.5 million dwarfs and 0.5 million giants stars) in the Milky Way, based on stellar colors from the Javalambre Photometric Local Universe Survey (J-PLUS) DR3 and \textit{Gaia} EDR3. These estimates are obtained through the construction of a large spectroscopic training set with parameters and abundances adjusted to uniform scales, and trained with a Kernel Principal Component Analysis. Owing to the seven narrow/medium-band filters employed by J-PLUS, we obtain precisions in the abundance estimates that are as good or better than derived from medium-resolution spectroscopy for stars covering a wide range of the parameter space: 0.10-0.20 dex for [Fe/H] and [C/Fe], and 0.05 dex for [Mg/Fe] and [$\alpha$/Fe]. Moreover, systematic errors due to the influence of molecular carbon bands on previous photometric-metallicity estimates (which only included two narrow/medium-band blue filters) have now been removed, resulting in photometric-metallicity estimates down to [Fe/H] $\sim -4.0$, with typical uncertainties of 0.25 dex and 0.40 dex for dwarfs and giants, respectively. This large photometric sample should prove useful for the exploration of the assembly and chemical-evolution history of our Galaxy., Comment: 31 pages, 20 figures, 3 tables, accepeted by ApJ

Mobile app review analysis presents unique challenges due to the low quality, subjective bias, and noisy content of user-generated documents. Extracting features from these reviews is essential for tasks such as feature prioritization and sentiment analysis, but it remains a challenging task. Meanwhile, encoder-only models based on the Transformer architecture have shown promising results for classification and information extraction tasks for multiple software engineering processes. This study explores the hypothesis that encoder-only large language models can enhance feature extraction from mobile app reviews. By leveraging crowdsourced annotations from an industrial context, we redefine feature extraction as a supervised token classification task. Our approach includes extending the pre-training of these models with a large corpus of user reviews to improve contextual understanding and employing instance selection techniques to optimize model fine-tuning. Empirical evaluations demonstrate that this method improves the precision and recall of extracted features and enhances performance efficiency. Key contributions include a novel approach to feature extraction, annotated datasets, extended pre-trained models, and an instance selection mechanism for cost-effective fine-tuning. This research provides practical methods and empirical evidence in applying large language models to natural language processing tasks within mobile app reviews, offering improved performance in feature extraction., Comment: 46 pages, 8 tables, 11 figures

Background: Machine learning (ML) model composition is a popular technique to mitigate shortcomings of a single ML model and to design more effective ML-enabled systems. While ensemble learning, i.e., forwarding the same request to several models and fusing their predictions, has been studied extensively for accuracy, we have insufficient knowledge about how to design energy-efficient ensembles. Objective: We therefore analyzed three types of design decisions for ensemble learning regarding a potential trade-off between accuracy and energy consumption: a) ensemble size, i.e., the number of models in the ensemble, b) fusion methods (majority voting vs. a meta-model), and c) partitioning methods (whole-dataset vs. subset-based training). Methods: By combining four popular ML algorithms for classification in different ensembles, we conducted a full factorial experiment with 11 ensembles x 4 datasets x 2 fusion methods x 2 partitioning methods (176 combinations). For each combination, we measured accuracy (F1-score) and energy consumption in J (for both training and inference). Results: While a larger ensemble size significantly increased energy consumption (size 2 ensembles consumed 37.49% less energy than size 3 ensembles, which in turn consumed 26.96% less energy than the size 4 ensembles), it did not significantly increase accuracy. Furthermore, majority voting outperformed meta-model fusion both in terms of accuracy (Cohen's d of 0.38) and energy consumption (Cohen's d of 0.92). Lastly, subset-based training led to significantly lower energy consumption (Cohen's d of 0.91), while training on the whole dataset did not increase accuracy significantly. Conclusions: From a Green AI perspective, we recommend designing ensembles of small size (2 or maximum 3 models), using subset-based training, majority voting, and energy-efficient ML algorithms like decision trees, Naive Bayes, or KNN., Comment: Currently under review at a journal

This work introduces GPTCast, a generative deep-learning method for ensemble nowcast of radar-based precipitation, inspired by advancements in large language models (LLMs). We employ a GPT model as a forecaster to learn spatiotemporal precipitation dynamics using tokenized radar images. The tokenizer is based on a Quantized Variational Autoencoder featuring a novel reconstruction loss tailored for the skewed distribution of precipitation that promotes faithful reconstruction of high rainfall rates. The approach produces realistic ensemble forecasts and provides probabilistic outputs with accurate uncertainty estimation. The model is trained without resorting to randomness, all variability is learned solely from the data and exposed by model at inference for ensemble generation. We train and test GPTCast using a 6-year radar dataset over the Emilia-Romagna region in Northern Italy, showing superior results compared to state-of-the-art ensemble extrapolation methods., Comment: 20 pages, 10 figures

The latest advancements in machine learning, specifically in foundation models, are revolutionizing the frontiers of existing software engineering (SE) processes. This is a bi-directional phenomona, where 1) software systems are now challenged to provide AI-enabled features to their users, and 2) AI is used to automate tasks within the software development lifecycle. In an era where sustainability is a pressing societal concern, our community needs to adopt a long-term plan enabling a conscious transformation that aligns with environmental sustainability values. In this paper, we reflect on the impact of adopting environmentally friendly practices to create AI-enabled software systems and make considerations on the environmental impact of using foundation models for software development., Comment: Accepted in SE 2030 - International Workshop on Software Engineering in 2030

We used the Javalambre Photometric Local Universe Survey (J-PLUS) DR2 photometry in twelve optical bands over 2176 deg2 to estimate the fraction of white dwarfs with presence of CaII H+K absorption along the cooling sequence. We compared the J-PLUS photometry against metal-free theoretical models to estimate the equivalent width in the J0395 passband of 10 nm centered at 395 nm (EW_J0395), a proxy to detect calcium absorption. A total of 4399 white dwarfs within 30000 > Teff > 5500 K and mass M > 0.45 Msun were analyzed. Their EW_J0395 distribution was modeled using two populations, corresponding to polluted and non-polluted systems, to estimate the fraction of calcium white dwarfs (f_Ca) as a function of Teff. The probability for each individual white dwarf of presenting calcium absorption, pca, was also computed. The comparison with both the measured Ca/He abundance and the metal pollution from spectroscopy shows that EW_J0395 correlates with the presence of calcium. The fraction of calcium white dwarfs increases from f_Ca = 0 at Teff = 13500 K to f_Ca = 0.15 at Teff = 5500 K. We compare our results with the fractions derived from the 40 pc spectroscopic sample and from SDSS spectra. The trend found in J-PLUS observations is also present in the 40 pc sample, however SDSS shows a deficit of metal-polluted objects at Teff < 12000 K. Finally, we found 39 white dwarfs with pca > 0.99. Twenty of them have spectra presented in previous studies, whereas we observed six additional targets. These 26 objects were all confirmed as metal-polluted systems. The J-PLUS optical data provide a robust statistical measurement for the presence of CaII H+K absorption in white dwarfs. We find a 15 +- 3 % increase in the fraction of calcium white dwarfs from Teff = 13500 K to 5500 K, which reflects their selection function in the optical from the total population of metal-polluted systems., Comment: Submitted to Astronomy and Astrophysics. Comments are welcome

Downscaling techniques are one of the most prominent applications of Deep Learning (DL) in Earth System Modeling. A robust DL downscaling model can generate high-resolution fields from coarse-scale numerical model simulations, saving the timely and resourceful applications of regional/local models. Additionally, generative DL models have the potential to provide uncertainty information, by generating ensemble-like scenario pools, a task that is computationally prohibitive for traditional numerical simulations. In this study, we apply a Latent Diffusion Model (LDM) to downscale ERA5 data over Italy up to a resolution of 2 km. The high-resolution target data consists of 2-m temperature and 10-m horizontal wind components from a dynamical downscaling performed with COSMO_CLM. Our goal is to demonstrate that recent advancements in generative modeling enable DL to deliver results comparable to those of numerical dynamical models, given the same input data, preserving the realism of fine-scale features and flow characteristics. A selection of predictors from ERA5 is used as input to the LDM, and a residual approach against a reference UNET is leveraged in applying the LDM. The performance of the generative LDM is compared with reference baselines of increasing complexity: quadratic interpolation of ERA5, a UNET, and a Generative Adversarial Network (GAN) built on the same reference UNET. Results highlight the improvements introduced by the LDM architecture and the residual approach over these baselines. The models are evaluated on a yearly test dataset, assessing the models' performance through deterministic metrics, spatial distribution of errors, and reconstruction of frequency and power spectra distributions., Comment: 24 pages, 14 figures

This paper is focused on the segregation of FGK dwarf and giant stars through narrow-band photometric data using the Spanish Virtual Observatory (SVO) Filter Profile Service and associated photometric tools. We selected spectra from the MILES, STELIB, and ELODIE stellar libraries, and used SVO photometric tools to derive the synthetic photometry in 15 J-PAS narrow filters, which were especially selected to cover spectral features sensitive to gravity changes. Using machine-learning techniques as the Gaussian mixture model and the support vector machine, we defined several criteria based on J-PAS colours to discriminate between dwarf and giant stars. We selected five colour-colour diagrams that presented the most promising separation between both samples. Our results show an overall accuracy in the studied sample of $\sim$0.97 for FGK stars, although a dependence on the luminosity type and the stellar effective temperature was found. We also defined a colour-temperature relation for dwarf stars with effective temperatures between 4\,000 and 7\,000\,K, which allows one to estimate the stellar effective temperature from four J-PAS filters ($J0450$, $J0510$, $J0550$, and $J0620$). Additionally, we extended the study to M-type giant and dwarf stars, achieving a similar accuracy to that for FGK stars., Comment: 19 pages, 18 figures, accepted for publication in A&A

We study the impact of black hole nuclear activity on both the global and radial star formation rate (SFR) profiles in X-ray-selected active galactic nuclei (AGN) in the field of miniJPAS, the precursor of the much wider J-PAS project. Our sample includes 32 AGN with z < 0.3 detected via the XMM-Newton and Chandra surveys. For comparison, we assembled a control sample of 71 star-forming (SF) galaxies with similar magnitudes, sizes, and redshifts. To derive the global properties of both the AGN and the control SF sample, we used CIGALE to fit the spectral energy distributions derived from the 56 narrowband and 4 broadband filters from miniJPAS. We find that AGN tend to reside in more massive galaxies than their SF counterparts. After matching samples based on stellar mass and comparing their SFRs and specific SFRs (sSFRs), no significant differences appear. This suggests that the presence of AGN does not strongly influence overall star formation. However, when we used miniJPAS as an integral field unit (IFU) to dissect galaxies along their position angle, a different picture emerges. We find that AGN tend to be more centrally concentrated in mass with respect to SF galaxies. Moreover, we find a suppression of the sSFR up to 1Re and then an enhancement beyond 1Re , strongly contrasting with the decreasing radial profile of sSFRs in SF galaxies. This could point to an inside-out quenching of AGN host galaxies. These findings suggest that the reason we do not see differences on a global scale is because star formation is suppressed in the central regions and enhanced in the outer regions of AGN host galaxies. While limited in terms of sample size, this work highlights the potential of the upcoming J-PAS as a wide-field low-resolution IFU for thousands of nearby galaxies and AGN., Comment: 21 pages, 20 figures, Accepted for Publication in A&A

With its 12 optical filters, the Javalambre-Photometric Local Universe Survey (J-PLUS) provides an unprecedented multicolor view of the local Universe. The third data release (DR3) covers 3,192 deg$^2$ and contains 47.4 million objects. However, the classification algorithms currently implemented in its pipeline are deterministic and based solely on the sources morphology. Our goal is classify the sources identified in the J-PLUS DR3 images into stars, quasi-stellar objects (QSOs), and galaxies. For this task, we present BANNJOS, a machine learning pipeline that uses Bayesian neural networks to provide the probability distribution function (PDF) of the classification. BANNJOS is trained on photometric, astrometric, and morphological data from J-PLUS DR3, Gaia DR3, and CatWISE2020, using over 1.2 million objects with spectroscopic classification from SDSS DR18, LAMOST DR9, DESI EDR, and Gaia DR3. Results are validated using $1.4 10^5$ objects and cross-checked against theoretical model predictions. BANNJOS outperforms all previous classifiers in terms of accuracy, precision, and completeness across the entire magnitude range. It delivers over 95% accuracy for objects brighter than $r = 21.5$ mag, and ~90% accuracy for those up to $r = 22$ mag, where J-PLUS completeness is < 25%. BANNJOS is also the first object classifier to provide the full probability distribution function (PDF) of the classification, enabling precise object selection for high purity or completeness, and for identifying objects with complex features, like active galactic nuclei with resolved host galaxies. BANNJOS has effectively classified J-PLUS sources into around 20 million galaxies, 1 million QSOs, and 26 million stars, with full PDFs for each, which allow for later refinement of the sample. The upcoming J-PAS survey, with its 56 color bands, will further enhance BANNJOS's ability to detail each source's nature., Comment: Submitted to Astronomy and Astrophysics. 29 pages, 23 figures, 4 tables

We present the second NuSTAR and XMM-Newton extragalactic survey of the JWST North Ecliptic Pole (NEP) Time-Domain Field (TDF). The first NuSTAR NEP-TDF survey (Zhao et al. 2021) had 681 ks total exposure time executed in NuSTAR cycle 5, in 2019 and 2020. This second survey, acquired from 2020 to 2022 in cycle 6, adds 880 ks of NuSTAR exposure time. The overall NuSTAR NEP-TDF survey is the most sensitive NuSTAR extragalactic survey to date, and a total of 60 sources were detected above the 95% reliability threshold. We constrain the hard X-ray number counts, logN-log S, down to 1.7 x 10$^{-14}$ erg cm$^{-2}$ s$^{-1}$ at 8-24 keV and detect an excess of hard X-ray sources at the faint end. About 47% of the NuSTAR-detected sources are heavily obscured (NH > 10$^{23}$ cm$^{-2}$), and 18+20% of the NuSTAR-detected sources are Compton-thick (N>10$^{24}$ cm$^{-2}$). These fractions are consistent with those measured in other NuSTAR surveys. Four sources presented >2$\sigma$ variability in the 3-year survey. In addition to NuSTAR, a total of 62 ks of XMM-Newton observations were taken during NuSTAR cycle 6. The XMM-Newton observations provide soft X-ray (0.5-10keV) coverage in the same field and enable more robust identification of the visible and infrared counterparts of the NuSTAR-detected sources. A total of 286 soft X-ray sources were detected, out of which 214 XMM-Newton sources have secure counterparts from multiwavelength catalogs., Comment: 37 pages, 27 figures, Accepted by ApJ

The rapidly evolving fields of Machine Learning (ML) and Artificial Intelligence have witnessed the emergence of platforms like Hugging Face (HF) as central hubs for model development and sharing. This experience report synthesizes insights from two comprehensive studies conducted on HF, focusing on carbon emissions and the evolutionary and maintenance aspects of ML models. Our objective is to provide a practical guide for future researchers embarking on mining software repository studies within the HF ecosystem to enhance the quality of these studies. We delve into the intricacies of the replication package used in our studies, highlighting the pivotal tools and methodologies that facilitated our analysis. Furthermore, we propose a nuanced stratified sampling strategy tailored for the diverse HF Hub dataset, ensuring a representative and comprehensive analytical approach. The report also introduces preliminary guidelines, transitioning from repository mining to cohort studies, to establish causality in repository mining studies, particularly within the ML model of HF context. This transition is inspired by existing frameworks and is adapted to suit the unique characteristics of the HF model ecosystem. Our report serves as a guiding framework for researchers, contributing to the responsible and sustainable advancement of ML, and fostering a deeper understanding of the broader implications of ML models., Comment: Accepted at the 2024 ACM/IEEE 1st International Workshop on Methodological Issues with Empirical Studies in Software Engineering (WSESE)

Low rank parity check (LRPC) codes are the rank-metric analogue of low density parity check codes. In this paper we investigate a sub-family of LRPC codes, which have a parity check matrix defined over a subspace $V_{\alpha,d}=\langle 1,\alpha, \ldots, \alpha^{d-1}\rangle_{\mathbb{F}_q}\subsetneq \mathbb{F}_{q^m}$, where $\mathbb{F}_{q^m}$ is the finite field of $q^m$ elements and $d$ is significantly smaller than $m $. These codes are named bounded-degree LRPC (BD-LRPC) codes and are the same as the standard LRPC codes of density $2$ when the degree $d=2$, while BD-LRPC codes of degree $d>2$ constitute a proper subset of LRPC codes of density $d$. Exploiting the particular structure of their parity check matrix, we show that the BD-LRPC codes of degree $d$ can uniquely correct errors of rank weight $r$ when $n-k \geq r + u$ for certain $u \geq 1$, in contrast to the condition $n-k\geq dr$ required for the standard LRPC codes, where $d\geq n/(n-k)$. This underscores the superior decoding capability of the proposed BD-LRPC codes. As the code length $n$ approaches infinity, when $n/m\rightarrow 0$, it is shown that $u$ can be chosen as a certain constant, which indicates that the BD-LRPC codes with a code rate of $R$ can be, with a high probability, uniquely decodable with the decoding radius $\rho=r/n$ approaching the Singleton bound $1-R$ for $n \to \infty$; and when $b= n/m$ is a constant, the BD-LRPC codes can have unique decoding radius $\rho = 1-R-\epsilon $ for a small $\epsilon$, which can easily lead to $\rho>(1-R)/2$ with properly chosen parameters., Comment: Currently under review

Extreme Emission Line Galaxies (EELGs) stand as remarkable objects due to their extremely metal poor environment and intense star formation. Considered as local analogues of high-redshift galaxies in the peak of their star-forming activity, they offer insights into conditions prevalent during the early Universe. Assessment of their stellar and gas properties is, therefore, of critical importance, which requires the assembly of a considerable sample, comprehending a broad redshift range. The Javalambre-Physics of the Accelerating Universe Astrophysical Survey (JPAS) plays a significant role in assembling such a sample, encompassing approximately 8000 deg2 and employing 54 narrow-band optical filters. The present work describes the development and subsequent application of the tools that will be employed in the forthcoming JPAS spectrophotometric data, allowing for the massive and automated characterization of EELGs that are expected to be identified. This fully automated pipeline (requiring only the object coordinates from users) constructs Spectral Energy Distributions (SEDs) by retrieving virtually all the available multi-wavelength photometric data archives, employs SED fitting tools and identifies optical emission lines. It was applied to the sample of extreme line emitters identified in the miniJPAS Survey, and its derived physical properties such as stellar mass and age, coupled with fundamental relations, mirror results obtained through spectral modeling of SDSS spectra. Thorough testing using galaxies with documented photometric measurements across different wavelengths confirmed the pipeline's accuracy, demonstrating its capability for automated analysis of sources with varying characteristics, spanning brightness, morphology, and redshifts. The modular nature of this pipeline facilitates any addition from the user., Comment: 15 pages, 9 Figures, accepted for publication in MNRAS

In the context of requirements engineering, relation extraction is the task of documenting the traceability between requirements artefacts. When dealing with textual requirements (i.e., requirements expressed using natural language), relation extraction becomes a cognitively challenging task, especially in terms of ambiguity and required effort from domain-experts. Hence, in highly-adaptive, large-scale environments, effective and efficient automated relation extraction using natural language processing techniques becomes essential. In this chapter, we present a comprehensive overview of natural language-based relation extraction from text-based requirements. We initially describe the fundamentals of requirements relations based on the most relevant literature in the field, including the most common requirements relations types. The core of the chapter is composed by two main sections: (i) natural language techniques for the identification and categorization of requirements relations (i.e., syntactic vs. semantic techniques), and (ii) information extraction methods for the task of relation extraction (i.e., retrieval-based vs. machine learning-based methods). We complement this analysis with the state-of-the-art challenges and the envisioned future research directions. Overall, this chapter aims at providing a clear perspective on the theoretical and practical fundamentals in the field of natural language-based relation extraction., Comment: This article will appear as a chapter in a book provisionally titled "Natural Language Processing for Requirements Engineering", to be published by Springer

Mobile app reviews are a large-scale data source for software-related knowledge generation activities, including software maintenance, evolution and feedback analysis. Effective extraction of features (i.e., functionalities or characteristics) from these reviews is key to support analysis on the acceptance of these features, identification of relevant new feature requests and prioritization of feature development, among others. Traditional methods focus on syntactic pattern-based approaches, typically context-agnostic, evaluated on a closed set of apps, difficult to replicate and limited to a reduced set and domain of apps. Meanwhile, the pervasiveness of Large Language Models (LLMs) based on the Transformer architecture in software engineering tasks lays the groundwork for empirical evaluation of the performance of these models to support feature extraction. In this study, we present T-FREX, a Transformer-based, fully automatic approach for mobile app review feature extraction. First, we collect a set of ground truth features from users in a real crowdsourced software recommendation platform and transfer them automatically into a dataset of app reviews. Then, we use this newly created dataset to fine-tune multiple LLMs on a named entity recognition task under different data configurations. We assess the performance of T-FREX with respect to this ground truth, and we complement our analysis by comparing T-FREX with a baseline method from the field. Finally, we assess the quality of new features predicted by T-FREX through an external human evaluation. Results show that T-FREX outperforms on average the traditional syntactic-based method, especially when discovering new features from a domain for which the model has been fine-tuned., Comment: Accepted at IEEE International Conference on Software Analysis, Evolution and Reengineering (SANER 2024). 12 pages (including references), 5 figures, 4 tables

We study the local environmental properties of 418 supernovae (SNe) of all types using data from the Javalambre Photometric Local Universe Survey (J-PLUS), which includes 5 broad- and 7 narrow-band imaging filters, using two independent analyses: 1) the Normalized Cumulative Rank (NCR) method, utilizing all 12 single bands along with five continuum-subtracted narrow-band emission and absorption bands, and 2) simple stellar population (SSP) synthesis, where we build spectral energy distributions (SED) of the surrounding SN environment using the 12 filters. Improvements over previous works include: (i) the extension of the NCR technique to other filters using a set of homogeneous data; (ii) a correction for extinction to all bands based on the relation between the g-i color and the color excess; and (iii) a correction for the [NII] line contamination that falls within the H$\alpha$ filter. All NCR distributions in the broad-band filters, tracing the overall light distribution in each galaxy, are similar to each other, being type Ia, II and IIb SNe are preferably located in redder environments than the other SN types. The radial distribution of the SNe shows that type IIb SNe seem to have a preference for occurring in the inner regions of galaxies. All core-collapse SN (CC) types are strongly correlated to the [OII] emission, which traces SFR. The NCR distributions of the Ca II triplet show a clear division between II/IIb/Ia and Ib/Ic/IIn subtypes, which is interpreted as a difference in the environmental metallicity. Regarding the SSP synthesis, we found that including the 7 J-PLUS narrow filters in the fitting process has a more significant effect for the CC SN environmental parameters than for SNe Ia, shifting their values towards more extinct, younger, and more star-forming environments, due to the presence of strong emission-lines and stellar absorptions in those narrow-bands., Comment: 18 pages, 12 figures, 8 tables. A&A accepted

Galaxy clusters are an essential tool to understand and constrain the cosmological parameters of our Universe. Thanks to its multi-band design, J-PAS offers a unique group and cluster detection window using precise photometric redshifts and sufficient depths. We produce galaxy cluster catalogues from the miniJPAS, which is a pathfinder survey for the wider J-PAS survey, using the PZWav algorithm. Relying only on photometric information, we provide optical mass tracers for the identified clusters, including richness, optical luminosity, and stellar mass. By reanalysing the Chandra mosaic of the AEGIS field, alongside the overlapping XMM-Newton observations, we produce an X-ray catalogue. The analysis reveals the possible presence of structures with masses of 4$\times 10^{13}$ M$_\odot$ at redshift 0.75, highlighting the depth of the survey. Comparing results with those from two other cluster catalogues, provided by AMICO and VT, we find $43$ common clusters with cluster centre offsets of 100$\pm$60 kpc and redshift differences below 0.001. We provide a comparison of the cluster catalogues with a catalogue of massive galaxies and report on the significance of cluster selection. In general, we are able to recover approximately 75$\%$ of the galaxies with $M^{\star} >$2$\times 10^{11}$ M$_\odot$. This study emphasises the potential of the J-PAS survey and the employed techniques down to the group scales., Comment: 15 pages, 11 figures, 5 tables. Submitted to A&A in December 19, 2023

User reviews published in mobile app repositories are essential for understanding user satisfaction and engagement within a specific market segment. Manual analysis of reviews is impractical due to the large data volume, and automated analysis faces challenges like data synthesis and reporting. This complicates the task for app providers in identifying patterns and significant events, especially in assessing the influence of competitor apps. Furthermore, review-based research is mostly limited to a single app or a single app provider, excluding potential competition analysis. Consequently, there is an open research challenge in leveraging user reviews to support cross-app analysis within a specific market segment. Following a case-study research method in the microblogging app market, we introduce an automatic, novel approach to support mobile app market analysis. Our approach leverages quantitative metrics and event detection techniques based on newly published user reviews. Significant events are proactively identified and summarized by comparing metric deviations with historical baseline indicators within the lifecycle of a mobile app. Results from our case study show empirical evidence of the detection of relevant events within the selected market segment, including software- or release-based events, contextual events and the emergence of new competitors.

We aim at developing a robust methodology for constraining the luminosity and stellar mass functions (LMFs) of galaxies by solely using data from multi-filter surveys and testing the potential of these techniques for determining the evolution of the miniJPAS LMFs up to $z\sim0.7$. Stellar mass and $B$-band luminosity for each of the miniJPAS galaxies are constrained using an updated version of the SED-fitting code MUFFIT, whose values are based on composite stellar population models and the probability distribution functions of the miniJPAS photometric redshifts. Galaxies are classified through the stellar mass versus rest-frame colour diagram corrected for extinction. Different stellar mass and luminosity completeness limits are set and parametrised as a function of redshift, for setting limits in our flux-limited sample ($r_\mathrm{SDSS}<22$). The miniJPAS LMFs are parametrised according to Schechter-like functions via a novel maximum likelihood method accounting for uncertainties, degeneracies, probabilities, completeness, and priors. Overall, our results point to a smooth evolution with redshift ($0.0510.7$)., Comment: 31 pages, 15 figures. Submitted to A&A

Hugging Face (HF) has established itself as a crucial platform for the development and sharing of machine learning (ML) models. This repository mining study, which delves into more than 380,000 models using data gathered via the HF Hub API, aims to explore the community engagement, evolution, and maintenance around models hosted on HF, aspects that have yet to be comprehensively explored in the literature. We first examine the overall growth and popularity of HF, uncovering trends in ML domains, framework usage, authors grouping and the evolution of tags and datasets used. Through text analysis of model card descriptions, we also seek to identify prevalent themes and insights within the developer community. Our investigation further extends to the maintenance aspects of models, where we evaluate the maintenance status of ML models, classify commit messages into various categories (corrective, perfective, and adaptive), analyze the evolution across development stages of commits metrics and introduce a new classification system that estimates the maintenance status of models based on multiple attributes. This study aims to provide valuable insights about ML model maintenance and evolution that could inform future model development strategies on platforms like HF., Comment: Accepted at the 2024 IEEE/ACM 21th International Conference on Mining Software Repositories (MSR)

We present a new method for obtaining photometric redshifts (photo-z) for sources observed by multiple photometric surveys using a combination (conflation) of the redshift probability distributions (PDZs) obtained independently from each survey. The conflation of the PDZs has several advantages over the usual method of modelling all the photometry together, including modularity, speed, and accuracy of the results. Using a sample of galaxies with narrow-band photometry in 56 bands from J-PAS and deeper grizy photometry from the Hyper-SuprimeCam Subaru Strategic program (HSC-SSP), we show that PDZ conflation significantly improves photo-z accuracy compared to fitting all the photometry or using a weighted average of point estimates. The improvement over J-PAS alone is particularly strong for i>22 sources, which have low signal-to-noise ratio in the J-PAS bands. For the entire i<22.5 sample, we obtain a 64% (45%) increase in the number of sources with redshift errors |Dz|<0.003, a factor 3.3 (1.9) decrease in the normalised median absolute deviation of the errors (sigma_NMAD), and a factor 3.2 (1.3) decrease in the outlier rate compared to J-PAS (HSC-SSP) alone. The photo-z accuracy gains from combining the PDZs of J-PAS with a deeper broadband survey such as HSC-SSP are equivalent to increasing the depth of J-PAS observations by ~1.2--1.5 magnitudes. These results demonstrate the potential of PDZ conflation and highlight the importance of including the full PDZs in photo-z catalogues., Comment: 15 pages, 16 figures, 1 table. Accepted for publication in A&A

We employ the corrected Gaia Early Data Release 3 (EDR3) photometric data and spectroscopic data from the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) DR7 to assemble a sample of approximately 0.25 million FGK dwarf photometric standard stars for the 12 J-PLUS filters using the Stellar Color Regression (SCR) method. We then independently validated the J-PLUS DR3 photometry, and uncovered significant systematic errors: up to 15 mmag in the results of Stellar Locus (SL) method, and up to 10 mmag mainly caused by magnitude-, color-, and extinction-dependent errors of the Gaia XP spectra with the Gaia BP/RP (XP) Synthetic Photometry (XPSP) method. We have also further developed the XPSP method using the corrected Gaia XP spectra by Huang et al. (2023) and applied it to the J-PLUS DR3 photometry. This resulted in an agreement of 1-5 mmag with the SCR method, and a two-fold improvement in the J-PLUS zero-point precision. Finally, the zero-point calibration for around 91% of the tiles within the LAMOST observation footprint is determined through the SCR method, with the remaining approximately 9% of tiles outside this footprint relying on the improved XPSP method. The re-calibrated J-PLUS DR3 photometric data establishes a solid data foundation for conducting research that depends on high-precision photometric calibration., Comment: ApJS accepted; 21 papes; 20 figures, see main results in Figures 5 and 12

With changing climatic conditions, we are already seeing an increase in extreme weather events and their secondary consequences, including landslides. Landslides threaten infrastructure, including roads, railways, buildings, and human life. Hazard-based spatial planning and early warning systems are cost-effective strategies to reduce the risk to society from landslides. However, these both rely on data from previous landslide events, which is often scarce. Many deep learning (DL) models have recently been applied for landside mapping using medium- to high-resolution satellite images as input. However, they often suffer from sensitivity problems, overfitting, and low mapping accuracy. This study addresses some of these limitations by using a diverse global landslide dataset, using different segmentation models, such as Unet, Linknet, PSP-Net, PAN, and DeepLab and based on their performances, building an ensemble model. The ensemble model achieved the highest F1-score (0.69) when combining both Sentinel-1 and Sentinel-2 bands, with the highest average improvement of 6.87 % when the ensemble size was 20. On the other hand, Sentinel-2 bands only performed very well, with an F1 score of 0.61 when the ensemble size is 20 with an improvement of 14.59 % when the ensemble size is 20. This result shows considerable potential in building a robust and reliable monitoring system based on changes in vegetation index dNDVI only., Comment: Author 1 and Author 2 contributed equally to this work

We present a list of quasar candidates including photometric redshift estimates from the miniJPAS Data Release constructed using SQUEzE. This work is based on machine-learning classification of photometric data of quasar candidates using SQUEzE. It has the advantage that its classification procedure can be explained to some extent, making it less of a `black box' when compared with other classifiers. Another key advantage is that using user-defined metrics means the user has more control over the classification. While SQUEzE was designed for spectroscopic data, here we adapt it for multi-band photometric data, i.e. we treat multiple narrow-band filters as very low-resolution spectra. We train our models using specialized mocks from Queiroz et al. (2022). We estimate our redshift precision using the normalized median absolute deviation, $\sigma_{\rm NMAD}$ applied to our test sample. Our test sample returns an $f_1$ score (effectively the purity and completeness) of 0.49 for quasars down to magnitude $r=24.3$ with $z\geq2.1$ and 0.24 for quasars with $z<2.1$. For high-z quasars, this goes up to 0.9 for $r<21.0$. We present two catalogues of quasar candidates including redshift estimates: 301 from point-like sources and 1049 when also including extended sources. We discuss the impact of including extended sources in our predictions (they are not included in the mocks), as well as the impact of changing the noise model of the mocks. We also give an explanation of SQUEzE reasoning. Our estimates for the redshift precision using the test sample indicate a $\sigma_{NMAD}=0.92\%$ for the entire sample, reduced to 0.81\% for $r<22.5$ and 0.74\% for $r<21.3$. Spectroscopic follow-up of the candidates is required in order to confirm the validity of our findings., Comment: Accepted in A&A 24 pages, 24 figures, 7 tables

Nowadays, AI-based systems have achieved outstanding results and have outperformed humans in different domains. However, the processes of training AI models and inferring from them require high computational resources, which pose a significant challenge in the current energy efficiency societal demand. To cope with this challenge, this research project paper describes the main vision, goals, and expected outcomes of the GAISSA project. The GAISSA project aims at providing data scientists and software engineers tool-supported, architecture-centric methods for the modelling and development of green AI-based systems. Although the project is in an initial stage, we describe the current research results, which illustrate the potential to achieve GAISSA objectives., Comment: Accepted for publication as full paper - 2023 49th Euromicro Conference Series on Software Engineering and Advanced Applications (SEAA)

We present the Lyman-$a$ (Lya) Luminosity Function (LF) at $2.0543.5. This work serves as a proof-of-concept for the results that can be obtained with the upcoming data releases of the J-PAS survey., Comment: 25 pages, 15 figures, Accepted for publication in A&A

Samples of galaxy clusters allow us to better understand the physics at play in galaxy formation and to constrain cosmological models once their mass, position (for clustering studies) and redshift are known. In this context, large optical data sets play a crucial role. We investigate the capabilities of the Javalambre-Physics of the Accelerating Universe Astrophysical Survey (J-PAS) in detecting and characterizing galaxy groups and clusters. We analyze the data of the miniJPAS survey, obtained with the JPAS-Pathfinder camera and covering $1$ deg$^2$ centered on the AEGIS field to the same depths and with the same 54 narrow band plus 2 broader band near-UV and near-IR filters anticipated for the full J-PAS survey. We use the Adaptive Matched Identifier of Clustered Objects (AMICO) to detect and characterize groups and clusters of galaxies down to $S/N=2.5$ in the redshift range $0.05

To raise awareness of the huge impact Deep Learning (DL) has on the environment, several works have tried to estimate the energy consumption and carbon footprint of DL-based systems across their life cycle. However, the estimations for energy consumption in the training stage usually rely on assumptions that have not been thoroughly tested. This study aims to move past these assumptions by leveraging the relationship between energy consumption and two relevant design decisions in DL training; model architecture, and training environment. To investigate these relationships, we collect multiple metrics related to energy efficiency and model correctness during the models' training. Then, we outline the trade-offs between the measured energy consumption and the models' correctness regarding model architecture, and their relationship with the training environment. Finally, we study the training's power consumption behavior and propose four new energy estimation methods. Our results show that selecting the proper model architecture and training environment can reduce energy consumption dramatically (up to 80.72%) at the cost of negligible decreases in correctness. Also, we find evidence that GPUs should scale with the models' computational complexity for better energy efficiency. Furthermore, we prove that current energy estimation methods are unreliable and propose alternatives 2x more precise., Comment: 32 pages, 11 figures, under review in ACM Transactions on Software Engineering and Methodology (TOSEM). This work is an extension of arXiv:2307.05520v3 [cs.LG]

The rise of machine learning (ML) systems has exacerbated their carbon footprint due to increased capabilities and model sizes. However, there is scarce knowledge on how the carbon footprint of ML models is actually measured, reported, and evaluated. In light of this, the paper aims to analyze the measurement of the carbon footprint of 1,417 ML models and associated datasets on Hugging Face, which is the most popular repository for pretrained ML models. The goal is to provide insights and recommendations on how to report and optimize the carbon efficiency of ML models. The study includes the first repository mining study on the Hugging Face Hub API on carbon emissions. This study seeks to answer two research questions: (1) how do ML model creators measure and report carbon emissions on Hugging Face Hub?, and (2) what aspects impact the carbon emissions of training ML models? The study yielded several key findings. These include a stalled proportion of carbon emissions-reporting models, a slight decrease in reported carbon footprint on Hugging Face over the past 2 years, and a continued dominance of NLP as the main application domain. Furthermore, the study uncovers correlations between carbon emissions and various attributes such as model size, dataset size, and ML application domains. These results highlight the need for software measurements to improve energy reporting practices and promote carbon-efficient model development within the Hugging Face community. In response to this issue, two classifications are proposed: one for categorizing models based on their carbon emission reporting practices and another for their carbon efficiency. The aim of these classification proposals is to foster transparency and sustainable model development within the ML community., Comment: Accepted at the 2023 ACM/IEEE International Symposium on Empirical Software Engineering and Measurement (ESEM)

Due to their weak algebraic structure, low rank parity check (LRPC) codes have been employed in several post-quantum cryptographic schemes. In this paper we propose new improved decoding algorithms for (n, k) LRPC codes of dual rank weight d. The proposed algorithms can efficiently decode LRPC codes with the parameters satisfying n - k = rd - c, where r is the dimension of the error support and c <= d - 2. They outperform the original decoding algorithm of LRPC codes when d > 2 and allow for decoding LRPC codes with a higher code rate and smaller values m.

In this paper we generalize the notion of low-rank parity check (LRPC) codes by introducing a bilinear product over F^m q based on a generic 3-tensor in Fq^mxmxm, where Fq is the finite field with q elements. The generalized LRPC codes are Fq -linear codes in general and a particular choice of the 3-tensor corresponds to the original Fqm -linear LRPC codes. For the generalized LRPC codes, we propose two probabilistic polynomial-time decoding algorithms by adapting the decoding method for LRPC codes and also show that the proposed algorithms have a decoding failure rate similar to that of decoding LRPC codes, Comment: A shorter version of this paper was presented in ITW 2023

Diffusion models have been widely adopted in image generation, producing higher-quality and more diverse samples than generative adversarial networks (GANs). We introduce a latent diffusion model (LDM) for precipitation nowcasting - short-term forecasting based on the latest observational data. The LDM is more stable and requires less computation to train than GANs, albeit with more computationally expensive generation. We benchmark it against the GAN-based Deep Generative Models of Rainfall (DGMR) and a statistical model, PySTEPS. The LDM produces more accurate precipitation predictions, while the comparisons are more mixed when predicting whether the precipitation exceeds predefined thresholds. The clearest advantage of the LDM is that it generates more diverse predictions than DGMR or PySTEPS. Rank distribution tests indicate that the distribution of samples from the LDM accurately reflects the uncertainty of the predictions. Thus, LDMs are promising for any applications where uncertainty quantification is important, such as weather and climate., Comment: 18 pages, 6 figures. Submitted for publication