Shift In Distribution Research Articles

Dynamic Multiphysics Simulation of the Load-Following Behavior in a Typical Pressurized Water Reactor Power Plant

Most Nuclear Power Plants (NPPs) are designed for baseload operations, maintaining a steady power output at 100%, except during planned maintenance and refueling. However, in countries like France, Slovakia, and Korea, where nuclear power is a major source of electricity, integrating nuclear energy with intermittent renewables is crucial for stable power generation. This integration necessitates daily power adjustments by NPPs in response to grid demands, a process known as a Load Follow Operation (LFO). Such a process introduces strong interdependencies between thermal–hydraulic and neutron–kinetic parameters, coupled with the three-dimensional movement of Control Element Assemblies (CEAs) and Xenon dynamics, which pose safety challenges due to shifts in core power distribution. To address these complexities, a multi-physics approach is employed using the multi-physics package RELAP5/3DKIN and implementing two strategies. The first strategy uses a mechanical shim, adjusting the reactor power exclusively through CEAs. The second strategy combines CEA movement with adjustments in soluble boron concentration. Both strategies are evaluated against axial offset and 3D power peaking safety limits to ensure compliance with operational safety requirements.

A multicontinental dataset of butterfly thermal physiological traits

Butterflies serve as key indicators of climate change impacts such as shifts in emergence timing and shifts in geographic range and distribution. However, the development of commonly used ecological forecasts based on butterfly physiological tolerance of temperature change has lagged behind that of other taxonomic groups. Here, we provide a series of related datasets comprising butterfly thermal physiological traits to enable such forecasts. We compiled data from the literature on butterfly heat and cold tolerance (critical thermal maxima and minima) for 117 species as well as heat resistance (knockdown time) for 45 species. We also present a new dataset comprising heat and cold tolerance and thermal sensitivity of metabolic rate of 28 common North American butterfly species. We envision these data to not only provide foundations for contemporary ecological forecasts of vulnerability to recent climate change, but also to aid in our understanding of butterfly ecology and evolution over historical timescales.

Preserving Fairness in AI under Domain Shift

Existing algorithms for ensuring fairness in AI use a single-shot training strategy, where an AI model is trained on an annotated training dataset with sensitive attributes and then fielded for utilization. This training strategy is effective in problems with stationary distributions, where both the training and testing data are drawn from the same distribution. However, it is vulnerable with respect to distributional shifts in the input space that may occur after the initial training phase. As a result, the time-dependent nature of data can introduce biases and performance degradation into the model predictions, even if the model is initially fair. Model retraining from scratch using a new annotated dataset is a naive solution that is expensive and time-consuming. We develop an algorithm to adapt a fair model to remain fair and generalizable under domain shift using solely new unannotated data points. We recast this learning setting as an unsupervised domain adaptation (UDA) problem. Our algorithm is based on updating the model such that the internal representation of data remains unbiased despite distributional shifts in the input space. We provide empirical validation on three common fairness datasets to show that the challenge exists in practical setting and to demonstrate the effectiveness of our algorithm.

Climate warming drives population trajectories of freshwater fish

Climate change has emerged as a key threat to biodiversity, leading to broad-scale shifts in distributions of marine and terrestrial species as they attempt to track thermally suitable habitat. By contrast, our understanding of climate responses of freshwater species is relatively undeveloped, limiting our knowledge of whether projected warming will lead to freshwater biodiversity loss. Here, we linked a multicontinental database of riverine fish population abundance time series collected from 1958 to 2019 to temperature data from the same period. Across the sampled localities, waters warmed by 0.21 °C per decade (annual maximum of monthly temperatures). We tested whether fish responded to this change by i) increasing abundance at the cooler poleward limit of species distributions-predicted if warming has opened new opportunities-and ii) decreasing abundance toward the equatorward limit of distributions-predicted if temperatures have exceeded tolerance thresholds. We found that observed population trends were consistent with both of these expected patterns from climatic warming and that the trends were more pronounced in time series covering the longer time periods of 30+ y. The responses consistent with climate change were most evident in species with larger body sizes, higher trophic levels, river-sea migratory behavior, and more widespread distributions. Moreover, positive abundance responses to warming were more likely at higher altitudes where conditions tend to be cooler. These findings indicate that projected future warming will likely lead to widespread shifts in riverine community structure, including abundance declines at the trailing edge of species distributions.

Trophic ecology of groundfishes in nearshore areas of the Gulf of Maine.

Ecosystem management requires an integrated understanding of ecological interactions. In the Gulf of Maine (GoM), trophic information pertaining to commercially important groundfishes and nearshore prey communities is lacking. We characterized nearshore communities and groundfish diets using data collected from nearshore surveys (864 bottom trawls and 3638 stomach samples of six groundfish species) conducted biannually (spring and fall) in Midcoast Maine and Penobscot Bay from 2012 to 2022. Groundfish diets were dominated by some of the most available nearshore prey (gadiform and clupeiform fishes and pandalid and crangonid shrimps). Shifts in relative prey availability over environmental gradients (e.g., depth and position along the coast), across seasons, and over years corresponded with parallel patterns in prey contributions to groundfish diets in specific predator-prey interactions. Negative trends in the relative availability and diet occurrence of signature GoM prey taxa (Northern shrimp Pandalus borealis, Atlantic herring Clupea harengus, and euphausiids) indicate that broader ecosystem changes, such as steady rises in water temperature and shifts in species distributions, are impacting nearshore trophic dynamics in the GoM. These observations provide timely information on mechanisms that underlie groundfish productivity and warrant inclusion of nearshore trophic dynamics in relevant ecosystem models.

Species Distribution Models in plant conservation science: a comprehensive review with a focus on Iran

This review article thoroughly examines the role of Species Distribution Models (SDMs) in plant conservation science, with a specific focus on applications within Iran. Commencing with an extensive methodological approach, involving an exhaustive search across reputable academic databases such as Scopus, Web of Science, and Google Scholar, the review synthesizes a comprehensive set of studies. It offers deep insights into SDM principles, challenges, and transformative applications. Addressing these challenges, the review explores contemporary data collection methods, including the use of remote sensing, drones, and citizen science, which enhance the precision and scope of SDMs. A detailed examination of various modelling algorithms and approaches, including MaxEnt, Random Forest, Bayesian models, and others, highlights their specific applications and contributions to plant conservation. The review also integrates climate change data and various scenarios into SDMs, showcasing case studies that illustrate SDMs' potential to predict shifts in plant distributions in response to changing climate conditions and overexploitation. Emphasizing the importance of spatial scale, the review discusses its critical impact on the accuracy of modelling and conservation planning. The article concludes by underlining the indispensable role of SDMs in advancing plant conservation efforts, offering tailored recommendations for researchers, policymakers, and conservation practitioners.

Disentangled Dynamic Graph Attention Network for Out-of-Distribution Sequential Recommendation

Sequential recommendation, leveraging user-item interaction histories to provide personalized and timely suggestions, has drawn significant research interest recently. With the power of exploiting spatio-temporal dynamics, Dynamic Graph Neural Networks (DyGNNs) show great potential in sequential recommendation by modeling the dynamic relationship between users and items. However, spatio-temporal distribution shifts naturally exist in out-of-distribution sequential recommendation, where both user-item relationships and temporal sequences demonstrate pattern shifts. The out-of-distribution scenarios may lead to the failure of existing DyGNNs in handling spatio-temporal distribution shifts in sequential recommendation, given that the patterns they exploit tend to be variant w.r.t labels under distribution shifts. In this article, we propose Disentangled Intervention-based Dynamic graph Attention networks with Invariance Promotion ( I-DIDA ) to handle spatio-temporal distribution shifts in sequential recommendation by discovering and utilizing invariant patterns , i.e., structures and features whose predictive abilities are stable across distribution shifts. Specifically, we first propose a disentangled spatio-temporal attention network to capture the variant and invariant patterns. By utilizing the disentangled patterns, we design a spatio-temporal intervention mechanism to create multiple interventional distributions and an environment inference module to infer the latent spatio-temporal environments, and minimize the invariance loss to leverage the invariant patterns with stable predictive abilities under distribution shifts. Extensive experiments demonstrate the superiority of our method over state-of-the-art sequential recommendation baselines under distribution shifts.

Impact of climate change on the distribution of insectivorous bats: Implications for small-scale farming in southern Mexico

Bats provide important ecosystem services for agriculture, such as pest control, a function that is particularly relevant for small-scale farmers. However, climate change is causing a decrease in bat populations. To assess the potential impacts of climate change on insectivorous bats and the implications on small-scale farming of indigenous communities in the Chiapas Highlands in southern Mexico we developed a three-step protocol: (1) projecting distribution shifts of insectivorous bats under climate change scenarios using non-dispersal and limited-full dispersal assumptions, (2) using official information to estimate the average economic value of conducting chemical pest control in crops at a state level, (3) surveying small-scale farmers to estimate the local economic value of pest control and determine how bats are perceived by small-scale farmers. Our models project shifts in bat species due to climate change. Given that new suitable climatic areas are also projected, if we assume a limited-full dispersal scenario, bats might not be as at risk by climate change, but shifting their distribution to more suitable habitats will probably affect the dynamic of the ecosystem service they provide. The official estimated value of chemical pest control is $15.15 USD/ha, while the estimated cost resulting from a hybrid survey with small-scale farmers was $47.53 USD/ha. The difference in cost could be related to an overuse of insecticides for pest control or an increase in price due to reduced accessibility. Sixty percent of surveyed farmers perceived a decline in bat populations, 68% were unaware of the benefits that bats provide to their crops, 51% believe that bats are mainly hematophagous, and 10% recognize that people harm or kill bats. A new approach including communicating small-scale farmers in their native languages the benefits that insectivorous bats provide along with a pest management strategy for the efficient use of insecticides needs to be implemented.

Cross-correlation of Luminous Red Galaxies with Machine Learning Selected Active Galactic Nuclei in HSC-SSP: Unobscured AGN Residing in More Massive Halos

Active galactic nuclei (AGN) are the signposts of black hole growth, and likely play an important role in galaxy evolution. An outstanding question is whether AGN of different spectral types indicate different evolutionary stages in the coevolution of black holes and galaxies. We present the angular correlation function between an AGN sample selected from Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP) optical photometry and Wide-field Infrared Survey Explorer mid-IR photometry and a luminous red galaxy (LRG) sample from HSC-SSP. We investigate AGN clustering strength as a function of luminosity and spectral features across three independent HSC fields totaling ∼600 deg2, for z ∈ 0.6 −1.2 and AGN with L 6 μm > 3 × 1044 erg s−1. There are ∼28,500 AGN and ∼1.5 million LRGs in our primary analysis. We determine the average halo mass for the full AGN sample (M h ≈ 1012.9 h −1 M ⊙), and note that it does not evolve significantly as a function of redshift (over this narrow range) or luminosity. We find that, on average, unobscured AGN (M h ≈ 1013.3 h −1 M ⊙) occupy ∼4.5× more massive halos than obscured AGN (M h ≈ 1012.6 h −1 M ⊙), at 5σ statistical significance using 1D uncertainties, and at 3σ using the full covariance matrix, suggesting a physical difference between unobscured and obscured AGN, beyond the line-of-sight viewing angle. Furthermore, we find evidence for a halo mass dependence on reddening level within the Type I AGN population, which could support the existence of a dust-obscured phase. However, we also find that quite small systematic shifts in the redshift distributions of the AGN sample could explain current and previously observed differences in M h .

EWMA control chart framework for efficient Maxwell quality characteristic monitoring: An application to the aerospace industry

The Maxwell distribution is extensively employed in statistical modeling to analyze data, notably in the fields of chemistry, astrophysics, demography, actuarial sciences, economics, industry, and engineering. Recently, the application of the Maxwell data generating process (DGP) has focused extensively on the domain of statistical control charts. The V chart and VSQ chart are often used to identify unexpected changes in the distributional shift of the Maxwell process. The VSQ chart offers considerably better performance than the V chart for detecting moderate to large shifts in the scale parameter. However, the VSQ chart adopts the fundamental structure of the Shewhart monitoring scheme and is insensitive to small alterations in the target parameter. We propose a new control chart, namely, the Maxwell exponentially weighted moving average (MXEWMA) chart, for improved monitoring of quality attributes that are assumed to conform to Maxwell data generation. The factors used to design the parameters of the proposed chart are computed at different false alarm probabilities and across various sample sizes. The effectiveness of the suggested scheme is considered in terms of the different features of the run length (RL) distribution, including the average, median and standard deviation. A comparative study of the MXEWMA chart with the existing VSQ chart was performed across various sample sizes. The comparative analysis showed that the MXEWMA chart is an effective alternative and performs well in detecting reasonably small shifts in the parameter. Simulated data are employed to describe the computational procedure of the MXEWMA scheme. The simulation analysis demonstrated that the MXEWMA chart outperforms the existing method in identifying slight changes in the studied parameters. A real dataset is also considered to support the theoretical part of the work.

Integrating single-cell multi-omics data through self-supervised clustering

Advances in single-cell sequencing techniques enable individual cells to be sequenced across multiple modalities simultaneously, such as transcriptomics, epigenomics, and proteomics. Integrating multi-omics single-cell data provides a deeper and more comprehensive vision of genomic mechanisms. Due to the huge distribution shift between modalities, current integration methods mostly align the modality through domain adaption or similar strategies. They achieved limited performance likely because the modalities are over-divergent. Here, we propose a novel single-cell multimodal fusion method, scFPN, to improve the learned embedding by a clustering strategy. Specifically, scFPN first embeds each modality data through a feature pyramid network with a modality-specific variational auto-encoder. The learned hierarchical embeddings are then fused and input into a dual self-supervision optimizing module for attracting similar cells and separating dissimilar cells. We conducted comprehensive experiments on recently produced six datasets from different sequencing platforms and demonstrated the superiority of scFPN over a variety of state-of-the-art methods. More importantly, scFPN showed bio-interpretability by marker enrichment analysis from de-noising and imputing raw profiles.

A Multi-scale Patch Mixer Network for Time Series Anomaly Detection

With the development of Internet of Things (IoT) technology, a large amount of data with temporal characteristics is collected and stored. How to efficiently and accurately identify anomalies from these data is a major challenge. At present, there are many problems in the application of anomaly detection, including non-stationary data, complex and difficult-to-collect anomalies, the need for real-time detection and the limitation of computing resources. But few methods can comprehensively consider these issues. To overcome these challenges, we propose a lightweight neural network, Multi-scale Patch Mixer Network (MP-MixerNet). It is mainly composed of a Mixer Block based on fully connected layer design, which contains a Temporal-Mixer and a Spatial-Mixer, and can simultaneously model the intra- and inter-series dependencies of multivariate time series. We also perform multi-scale patch segmentation based on frequency analysis, which helps the model extract robust features from multiple period views. In addition, we design an Input Stabilization module to help the model deal with data distribution shift. Experimental results on a public time series anomaly detection dataset show that we are able to achieve higher comprehensive performance with fewer parameters and inference time.

The formative time delay and electron avalanche number distributions for multielectron initiation of streamer breakdown

Electron avalanche statistics were mainly studied with a single electron initiation and there was no mathematical transition reported to the formative time delay distribution. This paper deals with the transition from the electron avalanche number distribution to the formative time delay distribution for multielectron initiation and streamer breakdown mechanism. The goal of the research is a detailed analysis of the formative time delay of electrical breakdown for different applications, such as spark counters, resistive plate chambers, gas-discharge switches, and high voltage systems with gas as an insulator. The electron avalanche number distribution for multielectron initiation is described by negative binomial distribution (NBD). Regarding the newly derived formative time delay distribution, when the number of initiating electrons k is small, the formative time delay distributions are asymmetric with the pronounced right tail. With increasing k, the formative time distributions shift to the shorter formative times and become narrower and higher, more symmetric and Gauss-like. As statistical tests show, for k≳20 the hypothesis that the formative time delay distributions are Gaussians cannot be rejected. As illustrations, the formative time delay distributions in neon and air at low reduced electric field are shown and briefly discussed. The distributions for methylal, ether and methane are presented in order to compare the estimated distribution parameters with available experimental data and to model the measured distributions.

Climate-driven distributional shifts in Chocó endemic birds of southwest Colombia

IntroductionClimate change poses a significant threat to bird communities, especially forest-dwelling and narrowly distributed species, which are expected to experience severe range contractions and higher extinction risks compared to widely distributed and open-area species. The Chocó region in southwestern Colombia, known for its rich bird endemism, is particularly vulnerable.MethodsWe analyzed potential distribution shifts for 27 endemic and near-endemic bird species in the Chocó region using eBird occurrence records and climate projections. We modeled species distributions under low and high greenhouse gas emission scenarios for 2050 and 2070, comparing these projected distributions to current forested and protected areas to evaluate future conservation needs.ResultsOur findings indicate that nearly all species are projected to lose climate-suitable areas under at least one future scenario, resulting in a regional decline in species richness. Changes in species richness are most pronounced near the Colombia-Ecuador border, suggesting a shift to higher elevations. Notably, the Scarlet-and-white Tanager (Chrysothlypis salmoni) is predicted to suffer the greatest losses in climate-suitable area, both within protected and forested regions.DiscussionThese results highlight the urgency of expanding the protected area network and conserving key forested regions to help species adapt to climate change. By providing projected distribution maps and potential range shifts, our study underscores the importance of modeling future distributions to support conservation strategies for at-risk species and the ecological services they provide in tropical montane regions.

From Manufacturing Decline to the Rise of Consultancy

Since the creation of Quadros de Pessoal (QP) in 1981, a longitudinal employer-employee dataset encompassing all Portuguese firms, four different industrial classification methodologies have been in place. These changes have hindered the ability to conduct long-term inter-industry studies on firm dynamics and structural change. This paper presents a framework for transforming the QP into a dataset with time-consistent two-digit industry codes and examines shifts in the firm-level growth rate distribution, net job creation and employment composition across sectors. Our industry-consistent trends confirm the contraction of the manufacturing sector, which coincided with the expansion of the services sector. While the increased business dynamism and net job creation in activities arising from the ICT revolution played a critical role, the technological paradigm shift in Portugal appears to be predominantly reflected in the vertical disintegration of firms and the subsequent substantial growth of administrative support services.

D4A: An efficient and effective defense across agnostic adversarial attacks

Recent studies show that Graph Neural Networks (GNNs) are vulnerable to structure adversarial attacks, which draws attention to adversarial defenses in graph data. Previous defenses designed heuristic defense strategies for specific attacks or graph properties, and are no longer sufficiently robust across all these attacks. To address this problem, we discuss the abnormal behaviors of GNNs in structure perturbations from a posterior distribution perspective. We suggest that the structural vulnerability of GNNs stems from their dependence on local graph smoothing, which can also lead to unfitting — a first-found phenomenon specific to the graph domain. We demonstrate that abnormal behaviors, except for unfitting, can attribute to a posterior distribution shift. To intrinsically prevent the occurrence of abnormal behaviors, we first propose smooth-less message passing to enhance the tolerance with respect to structure perturbations, while significantly mitigating the unfitting. We also propose the distribution shift constraint to restrict other abnormal behaviors of our model. Our approach is evaluated on six different datasets across over four kinds of attacks and compared to 11 representative baselines. The experimental results show that our method improves the defense performance across various attacks, and provides a great trade-off between accuracy and adversarial robustness.

Understanding the temporal variability and predictability of streamflow signatures in the Colorado River Basin

It is well established that streamflow regimes evolve over decadal time scales (i.e., low frequency) leading to long term shifts in distributions. Similar low frequency variations have also been documented in streamflow predictability. Here we explore connections between streamflow distribution attributes and predictability regimes in the Upper Colorado River Basin. We employ nonlinear dynamical time series analysis methods on streamflow timeseries covering the period 762 – 2019 for six locations in the basin. First, a wavelet spectral analysis is performed to obtain the quasi-periodic ‘signal’ of the streamflow. The wavelet analysis also provides the temporal variability of the variance of the signal time series. The signal time series is embedded in a D-dimensional space with appropriate lag to reconstruct the phase space of the dynamics – i.e. the attractor. Overall predictability is assessed by quantifying the average divergence trajectories in the phase space using Global Lyapunov Exponents and the temporal variability of predictability via the Local Lyapunov Exponents. Results show clear oscillations in streamflow predictability with periods of both high and low predictability occurring throughout the study period at all gauges. Comparing predictability timeseries across the stream gauges we find that general consistency in high and low predictability periods, although they do not perfectly align temporally. In general, higher (lower) predictability periods are characterized by lower (higher) streamflow variance. While there is not a clear relationship between streamflow magnitude and predictability in general, modern high predictability epochs are characterized by a slightly greater likelihood of dry years and lower likelihood of wet years than other epochs. These findings indicate the potential for statistically significant differences in streamflow signatures between high and low predictability periods. Exploring these fundings further with potential connections to large-scale climate can be helpful in exploiting them for skillful short and medium term flow projections.

Effi

We delve into the realm of offline meta-reinforcement learning (OMRL), a practical paradigm in the field of reinforcement learning that leverages offline data to adapt to new tasks. While prior approaches have not explored the utilization of context-based dynamical models to tackle OMRL problems, our research endeavors to fill this gap. Our investigation uncovers shortcomings in existing context-based methods, primarily related to distribution shifts during offline learning and challenges in establishing stable task representations. To address these issues, we formulate the problem as Hidden-Parameter MDPs and propose a framework for effective model adaptation using meta-models plus latent variables, which is inferred by the transformer-based system recognition module trained in an unsupervised fashion. Through extensive experimentation encompassing diverse simulated robotics and control tasks, we validate the efficacy of our approach and demonstrate its superior generalization ability compared to existing schemes, and explore multiple strategies for obtaining policies with personalized models. Our method achieves a model with reduced prediction error, outperforming previous methods in policy performance, and facilitating efficient adaptation when compared to prior dynamic model generalization methods and OMRL algorithms.

Invariant Probabilistic Prediction

Abstract In recent years, there has been a growing interest in statistical methods that exhibit robust performance under distribution changes between training and test data. While most of the related research focuses on point predictions with the squared error loss, this article turns the focus towards probabilistic predictions, which aim to comprehensively quantify the uncertainty of an outcome variable given covariates. Within a causality-inspired framework, we investigate the invariance and robustness of probabilistic predictions with respect to proper scoring rules. We show that arbitrary distribution shifts do not, in general, admit invariant and robust probabilistic predictions, in contrast to the setting of point prediction. We illustrate how to choose evaluation metrics and restrict the class of distribution shifts to allow for identifiability and invariance in the prototypical Gaussian heteroscedastic linear model. Motivated by these findings, we propose a method to yield invariant probabilistic predictions and study the consistency of the underlying parameters. Finally, we demonstrate the empirical performance of our proposed procedure via simulations and on single-cell data.

Rescue Stenting for Failed Mechanical Thrombectomy in Acute Basilar Artery Occlusions: Analysis of the PC-SEARCH Registry.

We aimed to investigate whether rescue stenting (RS) following failed mechanical thrombectomy (MT) for acute basilar artery occlusion (BAO) results in superior functional outcomes and enhanced safety compared to the natural history of failed mechanical thrombectomy (MT). This retrospective cohort study utilized data from the PC-SEARCH registry, encompassing eight high-volume centers in the United States and covering the period from 2015 to 2021. Patients with BAO who experienced failed MT (mTICI score of 0-2a after at least one attempt of clot retrieval) were categorized based on whether they received additional intervention with RS. The primary outcome was a shift analysis of the 90-day mRS. Multivariable logistic regression was employed to assess both efficacy and safety outcomes. Of a total of 444 patients, 119 presented failed MT and were included in the analysis. The RS group comprised 65 (14.6%) patients, while the control group consisted of 54 (12.2%) patients. After adjusting, the RS group showed a favorable shift in the overall 90-day mRS distribution (acOR=4.56; 95% CI 1.67-12.45; p=0.003) and higher rates of 90-day 0-3 mRS (RS: 44.6% vs. control: 18.5%, aOR=7.57; 95% CI 1.91-30.12; p = 0.004) compared to the control group. RS also showed lower rates of 90-day mortality (RS: 43.1% vs. control: 64.8%, aOR=0.27; 95% CI 0.09-0.80; p=0.018) and comparable rates of sICH (RS: 3.1% vs. control: 13%, aOR=0.31; 95% CI 0.05-1.95; p=0.214). Our study demonstrated that RS is associated with improved functional outcomes and reduced mortality in BAO patients presenting MT failure. Further randomized trials are needed to validate these findings.