Edge Of Distribution Research Articles

Calibrating Bayesian Generative Machine Learning for Bayesiamplification

Abstract Recently, combinations of generative and Bayesian deep learning have been introduced in particle physics for both fast detector simulation and inference tasks.
These neural networks aim to quantify the uncertainty on the generated distribution originating from limited training statistics.
The interpretation of a distribution-wide uncertainty however remains ill-defined.
We show a clear scheme for quantifying the calibration of Bayesian generative machine learning models.
For a Continuous Normalizing Flow applied to a low-dimensional toy example, we evaluate the calibration of Bayesian uncertainties from either a mean-field Gaussian weight posterior, or Monte Carlo sampling network weights, to gauge their behaviour on unsteady distribution edges. 
Well calibrated uncertainties can then be used to roughly estimate the number of uncorrelated truth samples that are equivalent to the generated sample and clearly indicate data amplification for smooth features of the distribution.

Gene Flow Between Populations With Highly Divergent Mitogenomes in the Australian Stingless Bee, Tetragonula hockingsi.

Coadaptation of mitochondrial and nuclear genes is essential for proper cellular function. When populations become isolated, theory predicts that they should maintain mito-nuclear coadaptation in each population, even as they diverge in genotype. Mito-nuclear incompatibilities may therefore arise when individuals from populations with divergent co-evolved mito-nuclear gene sets are re-united and hybridise, contributing to selection against inter-population hybrids and, potentially, to speciation. Here, we explored genetic divergence and gene flow between populations of a stingless bee (Tetragonula hockingsi) that have highly divergent mitogenomes. We identified three distinct populations across the species' 2500 km range on the east coast of Queensland (Australia): 'Cape York', 'Northern', and 'Southern'. The mitogenomes of each population showed > 12% pairwise nucleotide divergence from each other, and > 7% pairwise amino acid divergence. Based on nuclear SNPs from reduced representation sequencing, we identified at least two zones of gene flow between populations: a narrow natural zone between Northern and Southern populations (coinciding with a biogeographic barrier, the Burdekin Gap), and an artificial zone at the southern edge of the species' distribution, where Cape York, Northern, and Southern mito-lineages have been brought together in recent decades due to beekeeping. In the artificial hybrid zone, we also confirmed that males of all three mito-lineages were attracted to the mating aggregations of Southern queens, consistent with inter-population hybridisation. Populations of T. hockingsi thus appear to be in the 'grey zone' of the speciation continuum, having strong genetic differentiation but incomplete reproductive isolation. Among the nuclear SNPs most differentiated between Northern and Southern populations, several were associated with genes involved in mitochondrial function, consistent with populations having co-diverged mito-nuclear gene sets. Our observations suggest that coadapted sets of mitochondrial and nuclear genes unique to each population of T. hockingsi may play a role in maintaining population boundaries, though more study is needed to confirm the fitness costs of mito-nuclear incompatibilities in hybrid individuals.

Historic Land Use Modifies Impacts of Climate and Isolation in Rear Edge European Beech (Fagus sylvatica L.) Populations.

Legacies of human land use have the potential to impact demographic responses to climate. However, few studies have investigated the interactive effects of land use legacies and climate change on tree demography. The demographic performance of rear edge populations in particular is an important determinant of a species' long-term persistence. In this study, we investigated whether human land use legacies affect demographic responses to climate and population isolation in rear edge European beech populations (Fagus sylvatica L.) at the temperate-Mediterranean transition zone in the NE Iberian Peninsula. We utilised data from the Spanish Forest Inventory and generalised linear mixed models to compare the potential interactions across four different demographic rates (tree growth, survival probability, new adult recruitment and sapling recruitment). We found that the demographic rates were affected by the combination of land use legacies, climate and population isolation in different ways, which could potentially lead to complex shifts in future population dynamics under climate change. We identified that intense historic management either magnified negative relationships between tree demography and climate or population isolation, or reduced demographic performance in favourable climates to levels observed in unfavourable climates. Through either form of interaction, we found that intense historic forest management had a negative impact on tree demography, which has the potential to compromise future carbon stocks and long-term population viability. Overall, we show that disentangling human and environmental factors can enable us to better understand heterogeneous demographic performance across the rear edge of species distributions.

Modelling and simulation of cloud‐native‐based edge computing terminals for power distribution

AbstractThe introduction of cloud‐native technology has significantly changed the architecture of applications and the mechanism for the collaborative operation of components in power distribution edge computing terminals (PDECT). To develop an effective quantitative analysis tool for PDECT performance, the composition and characteristics of cloud‐native PDECT are studied, and the modelling and simulation of cloud‐native PDECT are proposed. Subsequently, modelling is implemented through the simulation software CloudSim, achieving the simulation of microservices, containers, declarative configuration, and container orchestration with the consideration of power distribution scenarios. Then, by the proposed simulation scenario module, various elements of the power distribution scenarios can be self‐defined. Finally, by demonstrating the principles and implementation mechanisms of the proposed modelling method and simulation tool, and comparing simulation results for different service time ranges, access devices, resource configurations of PDECT, request occurrence rates, and resource scheduling strategies, the validity and effectiveness of the proposed modelling method and simulation tool are verified.

RGP: Neural Network Pruning Through Regular Graph With Edges Swapping.

Deep learning technology has found a promising application in lightweight model design, for which pruning is an effective means of achieving a large reduction in both model parameters and float points operations (FLOPs). The existing neural network pruning methods mostly start from the consideration of the importance of model parameters and design parameter evaluation metrics to perform parameter pruning iteratively. These methods were not studied from the perspective of network model topology, so they might be effective but not efficient, and they require completely different pruning for different datasets. In this article, we study the graph structure of the neural network and propose a regular graph pruning (RGP) method to perform a one-shot neural network pruning. Specifically, we first generate a regular graph and set its node-degree values to meet the preset pruning ratio. Then, we reduce the average shortest path-length (ASPL) of the graph by swapping edges to obtain the optimal edge distribution. Finally, we map the obtained graph to a neural network structure to realize pruning. Our experiments demonstrate that the ASPL of the graph is negatively correlated with the classification accuracy of the neural network and that RGP has a strong precision retention capability with high parameter reduction (more than 90%) and FLOPs reduction (more than 90%) (the code for quick use and reproduction is available at https://github.com/Holidays1999/Neural-Network-Pruning-through-its-RegularGraph-Structure).

Climatic limitations of nearly endangered Juniperus rigida populations at their range edges in semiarid China

The treat of climate warming is expected to increase the frequency and severity of droughts and significantly affect tree growth and distribution ranges, particularly in regions without monsoons. The distribution limits of tree species are highly susceptible to climate change. A deeper understanding of how tree species respond to climate across their distribution ranges is crucial for accurately predicting forest dynamics in the face of climate warming. Using tree ring width data from 586 trees at 19 sites spanning the northernmost and southernmost limits of Juniperus rigida (near-threatened, IUCN: NT), we assessed the growth responses of J. rigida to climate at different distribution range limits. We observed significant differences in tree-growth variability between the southern and northern populations of J. rigida, characterized by an increasing growth trend at the southern range limit, but a decreasing trend at the northern limit since 1996. Moreover, tree growth at their northern limit was hindered by high temperatures and SPEI during summer months, whereas it was constrained by low temperatures in March and precipitation in June at the southern limit. Particularly noteworthy is the increased sensitivity of growth to SPEI and precipitation in recent decades (1996–2018) compared to previous decades (1961–1995), particularly at their northern distribution edge. We therefore conclude that the growth of J. rigida trees at the southern limit could benefit from climate warming, while increasing warming and drought could limit tree growth or even cause tree mortality at the northern range limits. These findings may serve as an early warning for the growth and survival of J. rigida trees in open forests at their northern distribution edges in the face of future climate warming.

Landscape genetics of spruce grouse at the trailing edge of the boreal forest

Populations on the periphery of a species range generally occupy marginal or less continuous habitat and can have greater extinction risk than populations in the range‐center. For climate‐sensitive populations on the trailing edge of a species' distribution (i.e. low latitude or low elevation), the effects of climate change are expected to compound these threats, increasing the vulnerability of these populations. The spruce grouse Canachites canadensis is a boreal forest specialist that is expected to experience range contraction at the trailing edge of its range due to climate change. In this study, we investigated the genetic structure, genetic diversity, and connectivity of spruce grouse in Minnesota, along the southeastern range limit in the United States. Clustering algorithms and genetic diversity indices suggested a single continuous population occurred throughout northern Minnesota. We detected no signatures of recent inbreeding or population bottlenecks. We used maximum likelihood population effects modeling and identified coniferous forest land cover and lower average spring temperatures as predictors of gene flow. We used two approaches based on circuit theory to map the potential for gene flow and identified regions of consistently high gene flow in the northwest and northeast of the study area. Omniscape, which incorporated additional parameters representing the likely origins of dispersers and the maximum dispersal distance, suggested a more continuous gene flow landscape than did Circuitscape. Our work highlights the current genetic integrity of one of the largest populations of the eastern subspecies of spruce grouse C. c. canadensis in the United States, identifies key landscape attributes for functional connectivity, and demonstrates complementary approaches of Circuitscape and Omniscape for gene flow mapping.

Direct and indirect effects of climate and seed dynamics on the breeding performance of a seed predator at the distribution edge.

Marginal populations usually have low densities and are considered to be particularly vulnerable to environmental stochasticity. Using data collected in nest boxes, we analyzed the breeding performance of the edible dormouse (Glis glis), an obligate hibernating rodent and a seed predator in deciduous forests, in two populations at the distribution range's edge. Despite being only 20 km apart from each other, Montseny is a large patch of mixed deciduous forests (oaks and beech), whereas Montnegre would be the harshest habitat, that is, a small, isolated patch with only oaks. First, we studied the differences in climate and tree cover change in the two populations. Second, we analyzed the direct and indirect roles of local climate conditions and seed availability on breeding performance over 10 years in each population. Finally, we explored the influence of tree cover change on the occupancy dynamics in the two populations. Our results showed contrasting responses between populations: in Montseny, asynchronous seed production between oaks and beech precluded skip breeding, and breeding performance increased with seed availability. Furthermore, dormice in Montseny may use pollen production to anticipate the amount of beech nut resources and adjust their breeding effort. Boxes showed higher occupancy and colonization and fewer extinctions in Montseny than in Montnegre, where seed availability did not drive breeding performance. Results from Montnegre suggest that skip breeding was an adaptive response to a more pulsed, harsher environment. Here, females produced a similar number of pups than at Montseny. Long-term studies dealing with population responses in marginal habitats can lead to a deeper understanding of the capacities of organisms to adapt to harsh environments. Although local adaptation is frequently documented across various taxa, studies at the distribution edge may shed light on our still limited comprehension of the underlying mechanisms responsible for its occurrence.

Range-Wide Assessment of Recent Longleaf Pine (Pinus palustris Mill.) Area and Regeneration Trends

Longleaf pine (Pinus palustris Mill.) is a conifer historically associated with an open forest ecosystem that extended across much of the coastal plain of the Southeastern United States. It now exists mainly in isolated fragments following the conversion of forests and the long-term disruption of the low-intensity fire regime upon which the species depends. Recent decades have seen efforts to restore longleaf pine forests by government and private landowners. This was reflected in analyses of national forest inventory data during two time periods, ca. 2009–2015 and 2016–2021, that showed increases in the estimated number of longleaf pine trees, the area of the longleaf pine forest type, and the number and area of planted longleaf pine, along with growth in mean plot-level longleaf pine carbon and importance value. At the same time, we found a decrease in the overall forest area containing longleaf pine, manifested across a variety of other forest types. These results point to a dynamic through which forests dominated by longleaf pine are becoming more widespread via restoration, while forests in which the species is a less important component are transitioning to other forest types or land uses. We also detected a decrease over time in the estimated number of longleaf seedlings across most states and forest types and a decline in naturally regenerated longleaf pine. To further assess regeneration trends in longleaf pine, we calculated the estimated proportion of small trees (seedlings and saplings) for the entire species and for seed zone sub-populations. We found a species-wide decrease in the proportion of small trees, from 82.1 percent to 75.1 percent. This reduction was most pronounced along the edges of the species distribution and could indicate less sustainable levels of regeneration in some areas. These results underscore the challenges of facilitating natural regeneration in this important species.

Internal response of vegetation growth to degrees of permafrost degradation in Northeast China from 2001 to 2020

ABSTRACT Ecosystems at the southern edge of the permafrost distribution are highly sensitive to global warming. Changes in soil freeze-thaw cycles can influence vegetation growth in permafrost regions. Extant studies mainly focused on analyzing the differences of vegetation dynamics in different permafrost regions. However, the intrinsic drivers of permafrost degradation on vegetation growth remain elusive yet. Based on the top temperature of permafrost (TTOP) model, we simulated the spatial distribution of permafrost in Northeast China (NEC) from 2001 to 2020. Using the data of the vegetation Net Primary Productivity (NPP), vegetation phenology, climate and permafrost phenology, and analytical methods including partial correlation, multiple linear regression, and path analysis, we explored the response of vegetation growth and phenology to soil freeze-thaw changes and climate change under different degrees of permafrost degradation. Overall, the start date of the growing season (SOS) was very sensitive to the start date of soil thaw (SOT) changes, and multiple regression analyses showed that SOT was the main factor influencing SOS in 41.8% of the NEC region. Climatic factors remain the main factors affecting vegetation NPP in NEC, and the results of partial correlation analysis showed that only 9.7% of the regional duration of soil thaw (DOT) had a strong correlation with vegetation NPP. Therefore, we determined the mechanism responsible for the soil freeze-thaw changes and vegetation growth relationship using the path analysis. The results indicated that there is a potential inhibitory effect of persistent permafrost degradation on vegetation growth. Our findings would contribute to the improvement of process-based models of forest dynamics in the boreal region, which would help to plan sustainable development and conservation strategies in permafrost areas.

Evaluating sustainability of Thalictrum petaloideum L., a peripheral population in Korea

Conservation of rare plants, both at the national level and the regional level is important to maintain and conserve plant diversity. This study focused on Thalictrum petaloideum L., an endangered species, for which the Korean peninsula represents the rear edge of the global distribution. T. petaloideum is distributed in isolated populations across four regions. Yeongwol, in Gangwon, had the largest extent of distribution (27,663 m2), the highest number of sub-populations (204), and the highest total population size (87,681), followed by Danyang, in Chungbuk (3970 m2/28/10,285), Samcheok, in Gangwon (3390 m2/13/3871), and Yeosu, in Jeolanam (800 m2/10/816), in descending order. The most stable population structure was found in Yeongwol. Yeosu was found to have the most unstable population structure and the highest risk of extinction. To ensure the sustainability of the population in Yeosu, artificial management of climbing plants, such as Pueraria lobata, is recommended to maintain spaces suitable for regeneration. In Yeongwol, the population is likely to be sustained over a long period. The maintenance of traditional graveyards is necessary to ensure the population’s sustainability. Further studies should assess the genetic characteristics of each regional population for an in-depth characterization of plant distribution on the Korean peninsula.

Using individualized structural covariance networks to analyze the heterogeneity of cerebral small vessel disease with cognitive impairment

BackgroundCerebral small vessel disease (CSVD) includes vascular disorders characterized by heterogeneous pathomechanisms and different neuropathological clinical manifestations. Cognitive dysfunction in CSVD is associated with reductions in structural covariance networks (SCNs). A majority of research conducted on SCNs focused on group-level analysis. However, it is crucial to investigate the individualized variations in order to gain a better understanding of heterogeneous disorders such as CSVD. Therefore, this study aimed to utilize individualized differential structural covariance network (IDSCN) analysis to detect individualized structural covariance aberration. MethodsA total of 35 healthy controls and 33 CSVD patients with cognitive impairment participated in this investigation. Using the regional gray matter volume in their T1 images, the IDSCN was constructed for each participant. Finally, the differential structural covariance edges between the two groups were determined by comparing their IDSCN using paired-sample t-tests. On the basis of these differential edges, the two subtypes of cognitively impaired CSVD patients were identified. ResultsThe findings revealed that the differential structural covariance edges in CSVD patients with cognitive impairment showed a highly heterogeneous distribution, with the edges primarily cross-distributed between the occipital lobe (specifically inferior occipital gyrus and cuneus), temporal lobe (specifically superior temporal gyrus), and the cerebellum. To varying degrees, the inferior frontal gyrus and the superior parietal gyrus were also distributed. Subsequently, a correlation analysis was performed between the resulting differential edges and the cognitive scale scores. A significant negative association was observed between the cognitive scores and the differential edges distributed in the inferior frontal gyrus and inferior occipital gyrus, the superior temporal gyrus and inferior occipital gyrus, and within the temporal lobe. Particularly in the cognitive domain of attention, the two subtypes separated by differential edges exhibited differences in cognitive scale scores [Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA)]. The differential edges of the subtype 1, characterized by lower cognitive level, were mainly cross-distributed in the limbic lobe (specifically the cingulate gyrus and hippocampus), the parietal lobe (including the superior parietal gyrus and precuneus), and the cerebellum. In contrast, the differential edges of the subtype 2 with a relatively high level of cognition were distributed between the cuneus and the cerebellum. ConclusionsThe differential structural covariance was investigated between the healthy controls and the CSVD patients with cognitive impairment, showing that differential structural covariance existed between the two groups. The edge distributions in certain parts of the brain, such as cerebellum and occipital and temporal lobes, verified this. Significant associations were seen between cognitive scale scores and some of those differential edges .The two subtypes that differed in both differential edges and cognitive levels were also identified. The differential edges of subtype 1 with relatively lower cognitive levels were more distributed in the cingulate gyrus, hippocampus, superior parietal gyrus, and precuneus. This could potentially offer significant benefits in terms of accurate diagnosis and targeted treatment of heterogeneous disorders such as CSVD.

Topographic and climatic effects on Pinus halepensis s.l. growth at its drought tolerance margins under climatic change

Under global warming, drought will reduce productivity of Pinus halepensis s.l. (subspecies halepensis and brutia) and cause a retreat from its rear edge distribution (latitudinal/elevational) in the Mediterranean. To test whether topography can influence this scenario, we studied for approximately 40 years the growth of six natural pine stands in water-limited habitats on the islands of Zakinthos and Samos (eastern Mediterranean Greece), and determined the critical moisture sources that drove pine growth. Dominant pines were selected with no permanent water sources under contrasting moisture conditions created by topography (“wet”-gulley/valley vs. “dry”-upslope habitats). The responses of P. halepensis s.l. to drought under a moderate and a worst case scenario were tested, projected under global warming (approx. − 25% and 40% in annual precipitation compared to 1961–1990 average). Our results show that “wet” habitat pines had higher productivity under normal to wet climate. However, the more precipitation declined, “wet” habitat tree growth was reduced at a significantly faster rate, but also showed a faster recovery, once rainfall returned. Thus, Pinus halepensis s.l. populations in gullies/valleys, may be more drought resilient and less likely to retreat towards higher elevation/latitudes under global warming, compared to pines on dry upslope sites. Under moderate drought, both ecosystems relied on deeper moisture pools supplied by rainfall of the previous 3–6 years (including the year of growth). However, valley/gully habitat pines on significantly deeper soils (and probably on deeper heavily weathered bedrock), appeared to utilize surface moisture from winter/spring rainfall more efficiently for survival and recovery. Thus, deep soils may provide the key “buffer” for pine survival in such ecosystems that could act as potential refugia for P. halepensis s.l. under climate change.

Phylogeography and bioclimatic models revealed a complicated genetic structure and future range shifts of Lymantria monacha L.

AbstractThe phylogeography of economically important forest pests is important for understanding their demographic and evolutionary history. Linking the genetic data obtained with the bioclimatic models helps reveal future demographic trends of the pest species studied. Lymantria monacha is a polyphagous species that feeds on numerous coniferous and deciduous trees throughout the Palaearctic and is known to cause catastrophic defoliation, particularly in Europe. In addition, data from various mapping programmes over the last decade have revealed changes in the distribution of L. monacha. Therefore, in this study, we decided to clarify the evolutionary and demographic history of this important forest species using genetic data complemented by bioclimatic modelling. Our results confirmed the systematic status and monophyly of L. monacha. However, the lack of a geographical pattern between the studied regions suggests that the current genetic structure may be the result of recent dispersal events. Moreover, we found that the areas of high genetic diversity are consistent with potential past range shifts and survival of changes in climate and host plant availability. These two main variables also seem to determine the future range of L. monacha. Also, our modelling confirmed a poleward shift in its range and with a significant retraction from its current southern edge of distribution.

Tree rings reveal the transient risk of extinction hidden inside climate envelope forecasts

Given the importance of climate in shaping species' geographic distributions, climate change poses an existential threat to biodiversity. Climate envelope modeling, the predominant approach used to quantify this threat, presumes that individuals in populations respond to climate variability and change according to species-level responses inferred from spatial occurrence data-such that individuals at the cool edge of a species' distribution should benefit from warming (the "leading edge"), whereas individuals at the warm edge should suffer (the "trailing edge"). Using 1,558 tree-ring time series of an aridland pine (Pinus edulis) collected at 977 locations across the species' distribution, we found that trees everywhere grow less in warmer-than-average and drier-than-average years. Ubiquitous negative temperature sensitivity indicates that individuals across the entire distribution should suffer with warming-the entire distribution is a trailing edge. Species-level responses to spatial climate variation are opposite in sign to individual-scale responses to time-varying climate for approximately half the species' distribution with respect to temperature and the majority of the species' distribution with respect to precipitation. These findings, added to evidence from the literature for scale-dependent climate responses in hundreds of species, suggest that correlative, equilibrium-based range forecasts may fail to accurately represent how individuals in populations will be impacted by changing climate. A scale-dependent view of the impact of climate change on biodiversity highlights the transient risk of extinction hidden inside climate envelope forecasts and the importance of evolution in rescuing species from extinction whenever local climate variability and change exceeds individual-scale climate tolerances.

Effectiveness of distributed stateless network server selection under strict latency constraints

We consider a set of network users (nodes), each generating latency-constrained service requests corresponding to the execution of computational tasks on servers positioned either within a cloud infrastructure or at the network edge. Within this framework, we systematically assess the efficacy of a distributed stateless server selection strategy, strategically performed by individual nodes. Leveraging principles from game theory, our study allows for a comparative analysis between the optimality achieved through globally orchestrated stateless allocation and a decentralized stateless server selection mechanism driven by the self-interested objectives of individual nodes.Our emphasis on stateless server allocation, rooted in a probabilistic selection framework between edge and cloud servers, stems from prior empirical revelations demonstrating the advantageous outcomes of determining the optimal distribution of edge and cloud tasks based on static network characteristics. Importantly, this determination occurs irrespective of the real-time network state.The suboptimal nature of the selfish allocation is quantified by the so called “price of anarchy,” a metric shown to approximate unity closely. This observation substantiates the justification for a distributed strategic implementation of stateless policies. This elucidation serves as a pivotal guide for crafting algorithms governing server selection, providing a quantitative validation of the efficacy inherent in distributed self-interested approaches.

CFL：Data distribution edge clustering algorithm based on deep Federated learning

Federated learning (FL) allows end-devices to train local models on their respective local datasets and collaborate with the server to train a global predictive model, thus achieving the goal of machine learning while protecting the privacy and sensitive data of end-devices. However, simultaneous access to the server by a large number of end devices may result in increased transmission latency and some local models may have malicious behavior, converging in the opposite direction to the global model. As a result, additional communication costs can occur during the federation learning process. Existing research has mainly focused on reducing the number of communication rounds or cleaning dirty local data. In order to decrease the overall amount of local updates, we provide an edge-based model cleaning and device clustering strategy in this study. By computing the cosine similarity between local update parameters and global model parameters over a multi-dimensional space, the approach assesses if local updates are necessary while preventing pointless communications. In addition, end devices are grouped together based on where they are connected to the network and connect to the cloud via mobile edge nodes in clusters, therefore lowering the latency associated with high concurrent server access. Convolutional neural networks and Soft max regression are used, for instance, to perform MNIST handwritten digit recognition, and the effectiveness of the suggested approach to enhancing communication efficiency is confirmed. According to experimental findings, the edge-based model cleaning and device clustering technique decreases the quantity of local updates by 60% and speeds up the model's convergence by 10.3% when compared to the conventional FL approach.

Boundary Effects Cause False Signals of Range Expansions in Population Genomic Data.

Studying range expansions is central for understanding genetic variation through space and time as well as for identifying refugia and biological invasions. Range expansions are characterized by serial founder events causing clines of decreasing genetic diversity away from the center of origin and asymmetries in the two-dimensional allele frequency spectra. These asymmetries, summarized by the directionality index (ψ), are sensitive to range expansions and persist for longer than clines in genetic diversity. In continuous and finite meta-populations, genetic drift tends to be stronger at the edges of the species distribution in equilibrium populations and populations undergoing range expansions alike. Such boundary effects are expected to affect geographic patterns in genetic diversity and ψ. Here we demonstrate that boundary effects cause high false positive rates in equilibrium meta-populations when testing for range expansions. In the simulations, the absolute value of ψ (|ψ|) in equilibrium data sets was proportional to the fixation index (FST). By fitting signatures of range expansions as a function of ɛ |ψ|/FST and geographic clines in ψ, strong evidence for range expansions could be detected in data from a recent rapid invasion of the cane toad, Rhinella marina, in Australia, but not in 28 previously published empirical data sets from Australian scincid lizards that were significant for the standard range expansion tests. Thus, while clinal variation in ψ is still the most sensitive statistic to range expansions, to detect true signatures of range expansions in natural populations, its magnitude needs to be considered in relation to the overall levels of genetic structuring in the data.

HFN: Heterogeneous feature network for multivariate time series anomaly detection

As the key step of anomaly detection for multivariate time-series (MTS) data, learning the relations among different variables has been explored by many approaches. However, most existing approaches overlook the heterogeneity among variables, that is, different types of variables (continuous numerical variables, discrete categorical variables or hybrid variables) may have different edge distributions. In this paper, we propose a novel semi-supervised anomaly detection framework based on a heterogeneous feature network (HFN) for MTS. Specifically, we first combine the embedding similarity subgraph generated by sensor embedding and the feature value similarity subgraph generated by sensor values to construct a time-series heterogeneous graph, which fully utilizes the rich heterogeneous mutual information among variables. Then, a prediction model containing nodes and channel attentions is jointly optimized to obtain better time-series representations. This approach fuses the state-of-the-art technologies of heterogeneous graph structure learning (HGSL) and representation learning. Experimental results on four sensor datasets from real-world applications demonstrate that our approach achieves more accurate anomaly detection compared to baseline methods, laying a foundation for the rapid positioning of anomalies.

A Zero-Shot Image Classification Method of Ship Coating Defects Based on IDATLWGAN

In recent years, the defect image classification method based on deep transfer learning has been widely explored and researched, and the task of source and target domains with the same painting defect image class has been solved successfully. However, in real applications, due to the complexity and uncertainty of ship painting conditions, it is very likely that there are unknown classes of painting defects, and the traditional deep learning model cannot identify a few classes, which leads to model overfitting and reduces its generalization ability. In this paper, a zero-shot Image classification method for ship painting defects based on IDATLWGAN is proposed to identify new unknown classes of defects in the target domain. The method is based on a deep convolutional neural network combined with adversarial transfer learning. First, a preprocessed ship painting defect dataset is used as input for the domain-invariant feature extractor. Then, the domain invariant feature extractor takes domain invariant features from the source and target domains. Finally, Defect discriminators and domain alignment discriminators are employed to classify the known categories of unlabeled defects and unknown categories of unlabeled defects in the target domain and to further reduce the distance between the edge distributions of the source and target domains. The experimental results show that the proposed model in this paper extracts a better distribution of invariant features in the source and target domains compared to other existing transfer learning models. It can successfully complete the migration task and accurately recognize the painting defects of known categories and new unknown categories, which is a perfect combination of intelligent algorithms and engineering practice.