Large Set Research Articles

Supporting Safety Analysis of Image-processing DNNs through Clustering-based Approaches

The adoption of deep neural networks (DNNs) in safety-critical contexts is often prevented by the lack of effective means to explain their results, especially when they are erroneous. In our previous work, we proposed a white-box approach (HUDD) and a black-box approach (SAFE) to automatically characterize DNN failures. They both identify clusters of similar images from a potentially large set of images leading to DNN failures. However, the analysis pipelines for HUDD and SAFE were instantiated in specific ways according to common practices, deferring the analysis of other pipelines to future work. In this paper, we report on an empirical evaluation of 99 different pipelines for root cause analysis of DNN failures. They combine transfer learning, autoencoders, heatmaps of neuron relevance, dimensionality reduction techniques, and different clustering algorithms. Our results show that the best pipeline combines transfer learning, DBSCAN, and UMAP. It leads to clusters almost exclusively capturing images of the same failure scenario, thus facilitating root cause analysis. Further, it generates distinct clusters for each root cause of failure, thus enabling engineers to detect all the unsafe scenarios. Interestingly, these results hold even for failure scenarios that are only observed in a small percentage of the failing images.

The history and nomenclatural significance of herbarium collections made by Alexander A. Tatarinow in North China and Mongolia in 1841–1850

AbstractAlexander A. Tatarinow (1817?–1886) made an extensive collection of vascular plants and insects in North China and Mongolia while serving as a physician in the 12th Russian Orthodox Ecclesiastical Mission in Beijing during 1841–1850. Tatarinow's plant collection included about 800 species and became the basis for 70 new species, of which 12 were named after him. The collection was split into three large sets, each labelled in its own way and handled in different times by different people at different institutions, thus obscuring its correct attribution; notable misinterpretations include its misattributions to A. von Bunge and P.E. Kirilov who were attached to the 11th Mission and collected plants near Beijing during 1830–1840. In this study, the history of these three sets and their parts is uncovered, and label samples are provided and commented to allow unambiguous recognition of the sets for correct lectotypification of plant species names based on the material collected by the 11th and 12th Missions.

Shock cooling emission from explosions of red supergiants: II. An analytic model of deviations from blackbody emission

ABSTRACT Light emission in the first hours and days following core-collapse supernovae (SNe) is dominated by the escape of photons from the expanding shock-heated envelope. In a preceding paper, Paper I, we provided a simple analytic description of the time-dependent luminosity, L, and colour temperature, Tcol, valid up to H recombination (T ≈ 0.7 eV), for explosions of red supergiants with convective polytropic envelopes without significant circumstellar medium (CSM). The analytic description was calibrated against ‘grey’ (frequency-independent) photon diffusion numeric calculations. Here, we present the results of a large set of 1D multigroup (frequency-dependent) calculations, for a wide range of progenitor parameters (mass, radius, core/envelope mass ratios, metalicity) and explosion energies, using opacity tables that we constructed (and made publicly available), including the contributions of bound–bound and bound–free transitions. We provide an analytic description of the small, ${\simeq}10\ \hbox{per cent}$ deviations of the spectrum from blackbody at low frequencies, hν < 3Tcol, and an improved (over Paper I) description of ‘line dampening’ for hν > 3Tcol. We show that the effects of deviations from initial polytropic density distribution are small, and so are the effects of ‘expansion opacity’ and deviations from LTE ionization and excitation (within our model assumptions). A recent study of a large set of type II SN observations finds that our model accounts well for the early multiband data of more than 50 per cent of observed SNe (the others are likely affected by thick CSM), enabling the inference of progenitor properties, explosion velocity, and relative extinction.

Enhancing 2-Pyrone Synthase Efficiency by High-Throughput Mass-Spectrometric Quantification and In Vitro/In Vivo Catalytic Performance Correlation.

Engineering efficient biocatalysts is essential for metabolic engineering to produce valuable bioproducts from renewable resources. However, due to the complexity of cellular metabolic networks, it is challenging to translate success in vitro into high performance in cells. To meet such a challenge, an accurate and efficient quantification method is necessary to screen a large set of mutants from complex cell culture and a careful correlation between the catalysis parameters in vitro and performance in cells is required. In this study, we employed a mass-spectrometry based high-throughput quantitative method to screen new mutants of 2-pyrone synthase (2PS) for triacetic acid lactone (TAL) biosynthesis through directed evolution in E. coli. From the process, we discovered two mutants with the highest improvement (46 fold) in titer and the fastest kcat (44 fold) over the wild type 2PS, respectively, among those reported in the literature. A careful examination of the correlation between intracellular substrate concentration, Michaelis-Menten parameters and TAL titer for these two mutants reveals that a fast reaction rate under limiting intracellular substrate concentrations is important for in-cell biocatalysis. Such properties can be tuned by protein engineering and synthetic biology to adopt these engineered proteins for the maximum activities in different intracellular environments.

A direct construction of cross z-complementary sequence sets with large set size

A direct construction of cross z-complementary sequence sets with large set size

An Example of Modelica–LabVIEW Communication Usage to Implement Hardware-in-the-Loop Experiments

Modelica is a very powerful language to simulate a very large set of systems, including electrical, thermal, mechanical, fluidic, control, and has already been used very extensively for several purposes, as the several Modelica conferences testify. Despite of this large literature, no paper seems to be available regarding the use of Modelica for real-time applications or hardware-in-the loop (HIL). This is a field where applications may be very fruitful. In this paper, the possibility of creating mixed software–hardware experiences (i.e., HIL), through combination of a Modelica program, the related simulation tool, a LabVIEW program, and the corresponding hardware is demonstrated. This demonstration is made using as an example a partial simulator of an electric vehicle running in a stand-alone PC, which communicates via User Datagram Protocol (UDP) packets with another PC running the LabVIEW program, which in turn is physically connected with the hardware-under-test. The obtained results are satisfying, given the inherent delay times due to the UDP communication.

A computational study of H-bonded networks in cyclic water clusters, (H2O)n (n = 3-12).

We have performed a detailed MM and DFT investigation of neutral water clusters (H2O)n (n = 3-12). Our results show the trend of interaction energies in these clusters as a function of the size of the cluster. They show that the H-bond strength increases with cluster size and that the model of water is better described if two different partial charges are used on the hydrogen, depending on whether hydrogen is H-bonded or not. The average binding enthalpy change due to the formation of H-bonds between water molecules is found to be - 25.9kJmol-1 at B3LYP/aug-cc-pVDZ level of theory. We observe the formation of cyclic H-bonded networks through the analysis of frontier orbitals and IR vibrational frequencies spectra. For the water cluster with n = 11, we observe an unusual reduction of the bandgap indicative of a cyclic H-bonded network. Calculations were performed with the MMFF94 force field and the B3LYP method using various large basis sets. Molecular orbital diagrams and population analysis were done using standard tools in Gaussian.

Associations of Lifestyle Patterns with Glucose and Lipid Metabolism in Finnish Adults at Increased Risk of Type 2 Diabetes.

Various lifestyle and sociodemographic factors have been associated with risk factors for type 2 diabetes (T2D). However, their combined associations with T2D risk factors have been studied much less. This study investigates cross-sectional associations of lifestyle patterns with T2D risk factors among 2925 adults at increased risk participating in the Stop Diabetes study. Lifestyle patterns are determined using principal component analysis (PCA) with several lifestyle and sociodemographic factors. The associations of lifestyle patterns with measures of glucose and lipid metabolism and serum metabolites analyzed by nuclear magnetic resonance (NMR) spectroscopy are studied using linear regression analysis. "Healthy eating" pattern is associated with better glucose and insulin metabolism, more favorable lipoprotein and fatty acid profiles and lower serum concentrations of metabolites related to inflammation, insulin resistance, and T2D. "High socioeconomic status and low physical activity" pattern is associated with increased serum concentrations of branched-chain amino acids, as are "Meat and poultry" and "Sleeping hours" patterns. "Snacks" pattern is associated with lower serum concentrations of ketone bodies. Our results show, in large scale primary care setting, that healthy eating is associated with better glucose and lipid metabolism and reveal novel associations of lifestyle patterns with metabolites related to glucose metabolism.

Differential amplification of the subtelomeric satellite DNA JcSAT1 in the genus Jatropha L. (Euphorbiaceae).

Satellite DNAs (satDNAs) are highly repetitive sequences that occur in virtually all eukaryotic genomes and can undergo rapid copy number and nucleotide sequence variation among relatives. After chromosomal mapping of the satDNA JcSAT1, it was found a large accumulation at subtelomeres of Jatropha curcas (subgenus Curcas), but an absence of these monomers in J. integerrima (subgenus Jatropha). This fact suggests a dynamic scenario for this satellite repeat in Jatropha genomes. Here, we used a multitasking approach (sequence analysis, DNA blotting and chromosomal mapping) to investigate the molecular organization and chromosomal abundance and distribution of JcSAT1 in a broader group of species from the subgenus Jatropha (J. gossypiifolia, J. mollissima, J. podagrica, and J. multifida) in addition to J. curcas, with the aiming of understanding the evolution of this satDNA. Based on the analysis of BAC clone sequences of J. curcas, a large array (~ 30kb) of 80 homogeneous monomers of JcSAT1 was identified in BAC 23J11. The monomer size was conserved (~ 358bp) and contained a telomeric motif at the 5' end. PCR amplification coupled with a Southern blot revealed the presence of JcSAT1-like sequences in all species examined. However, a large set of genome copies was identified only in J. curcas, where a ladder-like pattern with multimers of different sizes was observed. In situ hybridization of BAC 23J11 confirmed the subtelomeric pattern for J. curcas, but showed no signals on chromosomes of species from the subgenus Jatropha. Our data indicate that JcSAT1 is a highly homogeneous satDNA that originated from a region near the telomeres and spread throughout the chromosomal subtermini, possibly due to frequent ectopic recombination between these regions. The abundance of JcSAT1 in the genome of J. curcas suggests that an amplification event occurred either at the base of the subgenus Curcas or at least in this species, although the repeat is shared by all species of the genus studied so far.

Noisy image enhancements using deep learning techniques

This article explores the application of deep learning techniques to improve the accuracy of feature enhancements in noisy images. A multitasking convolutional neural network (CNN) learning model architecture has been proposed that is trained on a large set of annotated images. Various techniques have been used to process noisy images, including the use of data augmentation, the application of filters, and the use of image reconstruction techniques. As a result of the experiments, it was shown that the proposed model using deep learning methods significantly improves the accuracy of object recognition in noisy images. Compared to single-tasking models, the multi-tasking model showed the superiority of this approach in performing multiple tasks simultaneously and saving training time. This study confirms the effectiveness of using multitasking models using deep learning for object recognition in noisy images. The results obtained can be applied in various fields, including computer vision, robotics, automatic driving, and others, where accurate object recognition in noisy images is a critical component.

A Cultivar Attribute Database Derived from Ira J. Condit (1955): “Fig Varieties: A Monograph”

During California’s broad experiment with fig cultivation, Ira J. Condit became known as the “High Priest of the Fig.” Herein is a large set of fig attributes deemed agriculturally important by Ira J. Condit in his 1955 monograph. The attributes are divided into five categories and distributed across 717 cultivars. All data and software source code have been made available publicly on Figshare.

Zircon U–Pb ages of the Paleozoic volcaniclastic strata in the Junggar Basin, NW China

Abstract A large set of Paleozoic volcaniclastic rocks is exposed in the northwestern margin of the Junggar Basin from the southern part of the Central Asian Orogenic Belt. The Carboniferous volcaniclastic strata in this area have been studied in depth, and an accurate chronostratigraphic framework of these strata has been established. However, there is a lack of sufficient geochronological data for the deposition times of the other Paleozoic volcaniclastic strata. In this study, zircon U–Pb dating of the Ordovician, Silurian, and Devonian volcaniclastic strata in the area reveals that the youngest age of the tuffite sample collected from the originally defined Ordovician strata is 398 ± 11 Ma, which represents the age of volcanic activity during the period of tuffite deposition. Based on this finding, the originally defined Ordovician strata are redefined as the Lower Devonian. The youngest ages of the silty tuff samples collected from the originally defined Silurian strata peak are 445–418 Ma, so its age is Upper Silurian. The youngest ages of the tuffaceous sandstone samples collected from the originally defined Devonian strata peak are 346–342 Ma, so these Devonian are redefined as the Early Carboniferous strata. Two Archean ages (2,501 ± 12 and 3,193 ± 8 Ma) were obtained in Silurian strata, thus confirming the existence of metamorphic rock basement in the provenance areas from which the sediments were derived.

Topic Modeling: A Consistent Framework for Comparative Studies

In recent years, the field of Topic Modeling (TM) has grown in importance due to the increasing availability of digital text data. TM is an unsupervised learning technique that helps uncover latent semantic structures in large sets of documents, making it a valuable tool for finding relevant patterns. However, evaluating the performance of TM algorithms can be challenging as different metrics and datasets are often used, leading to inconsistent results. In addition, many current surveys of TM algorithms focus on a limited number of models and exclude state-of-the-art approaches. This paper has the objective of addressing these issues by presenting a comprehensive comparative study of five TM algorithms across three different benchmark datasets using five different metrics. We offer an updated survey of the latest TM approaches and evaluation metrics, providing a consistent framework for comparing different algorithms while introducing state-of-the art approaches that have been disregarded in the literature. The experiments, which primarily use Context Vectors (CV) Topic Coherence as an evaluation metric, show that Top2Vec is the best-performing model across all datasets, disrupting the tendency for Latent Dirichlet Allocation to be the best performer. Doi: 10.28991/ESJ-2024-08-01-09 Full Text: PDF

Pan-Cancer Interrogation of MUTYH Variants Reveals Biallelic Inactivation and Defective Base Excision Repair Across a Spectrum of Solid Tumors.

Biallelic germline pathogenic variants of the base excision repair (BER) pathway gene MUTYH predispose to colorectal cancer (CRC) and other cancers. The possible association of heterozygous variants with broader cancer susceptibility remains uncertain. This study investigated the prevalence and consequences of pathogenic MUTYH variants and MUTYH loss of heterozygosity (LOH) in a large pan-cancer analysis. Data from 354,366 solid tumor biopsies that were sequenced as part of routine clinical care were analyzed using a validated algorithm to distinguish germline from somatic MUTYH variants. Biallelic germline pathogenic MUTYH variants were identified in 119 tissue biopsies. Most were CRCs and showed increased tumor mutational burden (TMB) and a mutational signature consistent with defective BER (COSMIC Signature SBS18). Germline heterozygous pathogenic variants were identified in 5,991 biopsies and their prevalence was modestly elevated in some cancer types. About 12% of these cancers (738 samples: including adrenal gland cancers, pancreatic islet cell tumors, nonglioma CNS tumors, GI stromal tumors, and thyroid cancers) showed somatic LOH for MUTYH, higher rates of chromosome 1p loss (where MUTYH is located), elevated genomic LOH, and higher COSMIC SBS18 signature scores, consistent with BER deficiency. This analysis of MUTYH alterations in a large set of solid cancers suggests that in addition to the established role of biallelic pathogenic MUTYH variants in cancer predisposition, a broader range of cancers may possibly arise in MUTYH heterozygotes via a mechanism involving somatic LOH at the MUTYH locus and defective BER. However, the effect is modest and requires confirmation in additional studies before being clinically actionable.

Application of the Inclusion-Exclusion Principle to Prime Number Subsequences

We apply the Inclusion-Exclusion Principle to a unique pair of prime number subsequences to determine whether these subsequences form a small set or a large set and thus whether the infinite sum of the inverse of their terms converges or diverges. In this paper, we analyze the complementary prime number subsequences P′ and P′′ as well as revisit the twin prime subsequence P2.

The eROSITA upper limits

Context. The soft X-ray instrument eROSITA on board the Spectrum-Roentgen-Gamma (SRG) observatory has successfully completed four of the eight planned all-sky surveys, detecting almost one million X-ray sources during the first survey (eRASS1). The catalog of this survey will be released as part of the first eROSITA data release (DR1). Aims. Based on X-ray aperture photometry, we provide flux upper limits for eRASS1 in several energy bands. We cover galactic longitudes between 180° ≲ l ≲ 360° (eROSITA-DE). These data are crucial for studying the X-ray properties of variable and transient objects, as well as non-detected sources in the eROSITA all-sky survey data. Methods. We performed aperture photometry on every pixel of the SRG/eROSITA standard pipeline data products for all available sky tiles in the single detection band (0.2–2.3 keV). Simultaneously, we performed the same analysis in the three-band detection at soft (0.2–0.6 keV), medium (0.6–2.3 keV), and hard (2.3–5.0 keV) energy bands. Based on the combination of products for the individual bands, we are also able to provide aperture photometry products and flux upper limits for the 0.2–5.0 keV energy band. The upper limits were calculated based on a Bayesian approach that utilizes detected counts and background within the circular aperture. Results. The final data products consist of tables with the aperture photometry products (detected counts, background counts, and exposure time), a close-neighbor flag, and the upper flux limit based on an absorbed power-law spectral model (Γ = 2.0, NH = 3 × 1020 cm−2). The upper limits are calculated using the one-sided 3σ confidence interval (CL) of a normal distribution, representing CL = 99.87%. The aperture photometry products allow for an easy computation of upper limits at any other confidence interval and spectral model. These data are stored in a database with hierarchical indexes in order to offer a fast query option. Conclusions. We provide a detailed description of the process of retrieving SRG/eROSITA upper limits for a large set of input positions, as well as of the eROSITA data, the X-ray aperture photometry, the upper limit calculation, and the final data products. The eROSITA upper flux limits represent either the maximum flux of potential non-detections or the 3σ upper flux uncertainty of detected sources. We emphasize the importance of choosing the right spectral model that ought to match the spectral shape of the source of interest: the wrong spectral model can produce discrepancies of up to 30% in the final flux upper limit value. We also describe the architecture of the database and the web tool, which are designed to handle large queries on input positions.

Can We Boost N-Glycopeptide Identification Confidence? Smart Collision Energy Choice Taking into Account Structure and Search Engine.

High confidence and reproducibility are still challenges in bottom-up mass spectrometric N-glycopeptide identification. The collision energy used in the MS/MS measurements and the database search engine used to identify the species are perhaps the two most decisive factors. We investigated how the structural features of N-glycopeptides and the choice of the search engine influence the optimal collision energy, delivering the highest identification confidence. We carried out LC-MS/MS measurements using a series of collision energies on a large set of N-glycopeptides with both the glycan and peptide part varied and studied the behavior of Byonic, pGlyco, and GlycoQuest scores. We found that search engines show a range of behavior between peptide-centric and glycan-centric, which manifests itself already in the dependence of optimal collision energy on m/z. Using classical statistical and machine learning methods, we revealed that peptide hydrophobicity, glycan and peptide masses, and the number of mobile protons also have significant and search-engine-dependent influence, as opposed to a series of other parameters we probed. We envisioned an MS/MS workflow making a smart collision energy choice based on online available features such as the hydrophobicity (described by retention time) and glycan mass (potentially available from a scout MS/MS). Our assessment suggests that this workflow can lead to a significant gain (up to 100%) in the identification confidence, particularly for low-scoring hits close to the filtering limit, which has the potential to enhance reproducibility of N-glycopeptide analyses. Data are available via MassIVE (MSV000093110).

A fast and robust method for detecting trend turning points in InSAR displacement time series

Ground deformation monitoring is a crucial task in geohazard management to ensure the safety of lives and infrastructure. Persistent scatterer interferometric synthetic aperture radar (PS-InSAR) is an advanced technique for measuring small displacements on the Earth’s surface. Estimated PS-InSAR time series acquired by Sentinel-1 satellites provide a great opportunity for effective monitoring of ground deformation in recent years. However, challenges arise when processing these time series due to their non-uniform sampling, noise from atmosphere and preprocessing issues including phase unwrapping and others. Therefore, estimating the location and direction of trend turning in such time series, as an indicator of ground deformation, is not an easy task. In this work, a sequential turning point detection method (STPD) is proposed and compared with other change point detection methods. Using a large set of simulated time series with various noise types, it is shown that STPD outperforms other methods in terms of overall accuracy and root mean square error for location and direction of trend turnings. As a case study, STPD is applied to detect turning points within PS-InSAR time series for the province of Frosinone in Italy and classified using topography and land cover/use. In addition, an area susceptible to landslides is selected to estimate the starting dates of potential slow-moving landslides. It is also shown that the turning points in the local precipitation time series have a high correlation with the ones in the PS-InSAR time series, indicating that precipitation is a major triggering factor of the displacements in the area. The STPD can rapidly and effectively detect locations and directions of trend turnings and is freely available online in both MATLAB and python.

Parallel Processing of Image Databases for Accelerated Morphological Analysis

Nuclear forensics relies on different signatures to identify the source of nuclear material. Such signatures include crystalline structure, chemical composition, and particle morphology. One way to quantify morphology in electron microscope imagery is through image segmentation, where pixels are assigned to several partitions (or groups) that correspond to particles, grains, and other objects of interest within the image. Once pixels are assigned to segments, it is relatively straightforward to quantify other quantities of interest, such as grain size, circularity, etc. However, the range and diversity of microscope images make it difficult to obtain an accurate segmentation automatically. The accuracy of segmentation can be improved through supervised learning, but this requires many images to be manually segmented. Another way to improve the accuracy is to use interactive segmentation. Interactive segmentation requires a human to provide image-specific user input to improve performance. However, the amount of user input (effort) is generally far less than is required for supervised learning. In this paper, we investigate several parallelization strategies to automatically explore the user input parameter space of interactive segmentation algorithms across a large number of images. Specifically, we investigate four different parallelization mechanisms in a high-performance computing (HPC) environment and use the Amdahl fraction to evaluate efficiency on multiple processor cores across multiple nodes. Ultimately, the parallelization strategy that was most efficient utilized the message passing interface integrated with the segmentation and quantification code. This strategy had an Amdahl fraction of 0.985 and a performance of about 0.251 s/image. These results indicate that the parameter space of interactive segmentation algorithms can be efficiently explored using HPC. This opens the door to future work where user input is reduced and where interactive image segmentation algorithms are automatically applied to large image sets.

A comparison of compact, presumably young with extended, evolved radio active galactic nuclei

Context. The triggering and evolution of active galactic nuclei (AGNs) and the interaction of the AGN with its host galaxy is an important topic in extragalactic astrophysics. Radio sources with peaked spectra (peaked spectrum sources, PSS) and compact symmetric objects (CSO) are powerful, compact, and presumably young AGNs and therefore particularly suitable to study aspects of the AGN-host connection. Aims. We use a statistical approach to investigate properties of a PSS-CSO sample that are related to host galaxies and could potentially shed light on the link between host galaxies and AGNs. The main goal is to compare the PSS-CSO sample with a matching comparison sample of extended sources (ECS) to see if the two have significant differences. Methods. We analysed composite spectra, diagnostic line diagrams, multi-band spectral energy distributions (MBSEDs), star formation (SF) indicators, morphologies, and cluster environments for a sample of 121 PSSs and CSOs for which spectra are available from the Sloan Digital Sky Survey (SDSS). The statistical results were compared with those of the ECS sample, where we generally considered the two subsamples of quasi-stellar objects (QSOs) and radio galaxies separately. The analysis is based on a large set of archival data in the spectral range from the ultraviolet to mid-infrared. Results. We find significant differences between the PSS-CSO and the ECS sample. In particular, we find that the ECS sample has a higher proportion of passive galaxies with a lower star formation activity. This applies to both sub-samples (QSOs or radio galaxies) as well as to the entire sample. The star formation rates of the PSS-CSO host galaxies with available data are typically in the range ∼0 to 5 ℳ⊙ yr−1, and the stellar masses are in the range 3 × 1011 to 1012 ℳ⊙. Secondly, in agreement with previous results, we find a remarkably high proportion of PSS-CSO host galaxies with merger signatures. The merger fraction of the PSS-CSO sample is 0.61 ± 0.07, which is significantly higher than that of the comparison sample (0.15 ± 0.06). We suggest that this difference can be explained by assuming that the majority of the PSSs and CSOs cannot evolve to extended radio sources and are therefore not represented in our comparison sample.