Article
Computer Science, Artificial Intelligence
Hampus Linander, Oleksandr Balabanov, Henry Yang, Bernhard Mehlig
Summary: Bayesian inference can quantify uncertainty in neural network predictions using posterior distributions, and we show how prediction accuracy is related to epistemic and aleatoric uncertainties. We also introduce a novel acquisition function that outperforms common methods.
MACHINE LEARNING-SCIENCE AND TECHNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Kohei Ichikawa, Asaki Kataoka
Summary: This study trained two types of artificial neural networks, feedforward neural network (FFNN) and recurrent neural network (RNN), to perform sampling-based probabilistic inference. It found that the sampling mechanism in RNN efficiently utilizes the properties of dynamical systems, unlike FFNN. Additionally, the study found that sampling in RNNs provides an inductive bias and enables more accurate estimation than maximum a posteriori estimation.
NEURAL COMPUTATION
(2022)
Article
Environmental Sciences
Victor Hertel, Candace Chow, Omar Wani, Marc Wieland, Sandro Martinis
Summary: Geospatial resources, including SAR and optical data, play a crucial role in providing timely information about natural hazard events, such as floods, to support emergency response and hazard management efforts. However, accurate flood detection using SAR data is challenging, and unreliable information can lead to poor decision-making and negative consequences. Reliable uncertainty quantification is important in addressing this risk.
REMOTE SENSING OF ENVIRONMENT
(2023)
Article
Economics
Wei Xiao
Summary: This article proposes a simple heuristic to explain how people estimate the probability of an event with binary outcomes, such as a boom vs. a recession. The heuristic is based on base-rate neglect, a psychological trait where people tend to ignore prior probabilities and focus on more salient diagnostic information. An algorithm is constructed to replicate this behavior and its predictions match key features of the Anxious Index and stock market expectations. If investors use this heuristic in a consumption-based asset pricing model, the equity premium and its volatility will be higher compared to a rational expectations model.
JOURNAL OF ECONOMIC DYNAMICS & CONTROL
(2022)
Article
Computer Science, Artificial Intelligence
Badri N. Patro, Anupriy, Vinay P. Namboodiri
Summary: This paper proposes a probabilistic framework for solving the task of 'Visual Dialog', aiming to understand and analyze the sources of uncertainty for solving this task. The proposed probabilistic framework leads to an improved and more explainable visual dialog system.
PATTERN RECOGNITION
(2021)
Article
Management
Yanwei Jia, Jussi Keppo, Ville Satopaa
Summary: Experts' forecasts can be biased, especially due to herding bias. In certain scenarios, disclosure of public information may decrease forecasting accuracy, resulting in overly similar predictions and inflated variance among experts. However, the negative externality of public information no longer holds for probabilistic forecasts, which provide additional insights into experts' beliefs and interpersonal structure.
MANAGEMENT SCIENCE
(2023)
Article
Management
Emanuele Borgonovo, Gordon B. Hazen, Victor Richmond R. Jose, Elmar Plischke
Summary: This study focuses on selecting the most appropriate sensitivity measures for a specific reporting context. It proposes that the importance rankings of model inputs should correspond to their information value in constructing an optimal report, based on a proper scoring rule. The research examines the general conditions under which a sensitivity measure has this property and analyzes if sensitivity measures retain important properties like transformation invariance and compliance with Renyi's Postulate D for measures of statistical dependence.
EUROPEAN JOURNAL OF OPERATIONAL RESEARCH
(2021)
Article
Computer Science, Artificial Intelligence
Niklas Toetsch, Daniel Hoffmann
Summary: This study introduces a method to quantify the uncertainty of classification performance metrics based on a probability model of the confusion matrix. Results show that uncertainties can be surprisingly large limiting performance evaluation, while the method can also be used for sample size estimation.
PEERJ COMPUTER SCIENCE
(2021)
Article
Engineering, Industrial
Rafael Ballester-Ripoll, Manuele Leonelli
Summary: This paper shows how to apply Sobol's method of global sensitivity analysis to measure the influence of a set of nodes' evidence on a quantity of interest in a Bayesian network. The proposed method exploits the network structure to transform the problem of Sobol index estimation and gives exact results when exact inference is used. It also handles correlated inputs efficiently as long as eliminating the inputs' ancestors is computationally affordable.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2022)
Article
Physics, Multidisciplinary
Adam Thor Thorgeirsson, Frank Gauterin
Summary: The text discusses the importance of probabilistic predictions in machine learning applications and the challenges with Bayesian learning methods in terms of computational cost. It introduces a method to incorporate predictive uncertainty in federated learning by treating local weights as a posterior distribution for global model weights. By comparing this approach with state-of-the-art Bayesian and non-Bayesian algorithms, it demonstrates similar performance to benchmarks in a non-distributed setting when evaluated with proper scoring rules.
Article
Engineering, Industrial
Xian-Xun Yuan, Eishiro Higo, Mahesh D. Pandey
Summary: This paper presents a model to quantify the economic value gained by implementation of an inspection and preventive maintenance program, emphasizing the intricate interaction between parameter and temporal uncertainties associated with the degradation process. The research shows that the economic value is significantly sensitive to the prior information and relative costs of preventive and corrective maintenance.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2021)
Article
Engineering, Mechanical
Pengfei Wei, Fuchao Liu, Marcos Valdebenito, Michael Beer
Summary: Efficient propagation of imprecise probability models is achieved through the development of a new methodology framework named NIPI, focusing on the distributional probability-box model and the estimation of probabilistic moments of model responses. By integrating spatial correlation information revealed by the GPR model, NIPI estimations with high accuracy are derived, and numerical errors are treated as epistemic uncertainty.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2021)
Article
Mathematics, Applied
Avshalom Offner, Sam Manger, Jacques Vanneste
Summary: Positron emission particle tracking (PEPT) is an imaging method that reconstructs three-dimensional trajectories of small tracer particles in opaque fluids. This study introduces a probabilistic framework based on Bayesian inference for accurately determining particle positions from PEPT data, while accounting for scattering and noise. The formulation is illustrated with simulations, showing the optimal choice of observation time and the benefits of inferring particle velocity in addition to position.
Article
Engineering, Industrial
Krzysztof Woloszyk, Yordan Garbatov
Summary: This paper proposes a probabilistic-driven framework for enhanced corrosion estimation of ship structural components using Bayesian inference and limited measurement data. The framework incorporates measurement uncertainty and provides confidence intervals for the mean value and standard deviation.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2024)
Article
Engineering, Mechanical
Lechang Yang, Sifeng Bi, Matthias G. R. Faes, Matteo Broggi, Michael Beer
Summary: In this paper, a novel entropy-based metric utilizing Jensen-Shannon divergence is proposed to address inverse problems with mixed uncertainty, showing effectiveness and efficiency. By employing a discretized binning algorithm to reduce computation cost, the method demonstrates promising results in both static and dynamic systems.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2022)
Article
Evolutionary Biology
Cecilia Kardum Hjort, Josephine R. Paris, Peter Olsson, Lina Herbertsson, Joachim R. Miranda, Rachael Y. Dudaniec, Henrik G. Smith
Summary: The global movement of bees for agricultural pollination services can affect the local pollinator populations. This study investigates the genetic introgression and evolutionary divergence between wild and commercial bumblebees by comparing their whole genomes. The results show no widespread recent introgression of commercial bumblebees into local wild conspecific populations, but a highly divergent region on chromosome 11 in commercial bumblebees indicates different evolutionary processes compared to wild bumblebees.
EVOLUTIONARY APPLICATIONS
(2022)
Article
Biodiversity Conservation
Goran Andersson, Ted Proschwitz, Christoffer Fagerstrom, Martin Green, Henrik G. Smith, Ake Lindstrom
Summary: This study monitored the abundance of arthropods in a subalpine birch forest in Swedish Lapland over a period of 53 years. The results showed that there was no significant change in the arthropod numbers and biomass in this relatively unaffected area. This suggests that the factors causing arthropod declines identified in other habitats may have little to no impact due to the low human population density in this region.
INSECT CONSERVATION AND DIVERSITY
(2022)
Article
Ecology
Bjorn K. Klatt, Bethany Pudifoot, Pablo Urrutia-Cordero, Henrik G. Smith, Christian M. Alsterberg
Summary: Trophic cascades in the aquatic environment can lead to unexpected ecological interactions across the aquatic-terrestrial interface, facilitated by climate change. This study found that an aquatic trophic cascade induced the formation of algal surface mats, which bumblebees used as a water source during a heat wave and drought. This access to water was associated with higher bumblebee colony reproductive success, growth, and weight.
Article
Ecology
Nikolaos Alexandridis, Glenn Marion, Rebecca Chaplin-Kramer, Matteo Dainese, Johan Ekroos, Heather Grab, Mattias Jonsson, Daniel S. Karp, Carsten Meyer, Megan E. O'Rourke, Mikael Pontarp, Katja Poveda, Ralf Seppelt, Henrik G. Smith, Richard J. Walters, Yann Clough, Emily A. Martin
Summary: Controlling crop pests through adjusting the availability of host plants and natural enemy activity at a landscape scale can enhance agricultural sustainability. However, achieving natural pest control across different agroecological contexts requires a better understanding of its benefits. By combining trait-mediated understanding with mechanistic modeling, we can utilize existing empirical, theoretical, and methodological knowledge to improve predictions and management of natural pest control.
ECOLOGICAL APPLICATIONS
(2022)
Article
Ecology
Liam K. Kendall, John M. Mola, Zachary M. Portman, Daniel P. Cariveau, Henrik G. Smith, Ignasi Bartomeus
Summary: The size and sociality of species have an effect on their potential and realized foraging ranges. Larger body size corresponds to larger potential and realized ranges. Highly eusocial species have larger realized foraging ranges than primitively eusocial or solitary taxa.
Article
Biodiversity Conservation
Jessica L. Knapp, Adam Bates, Ove Jonsson, Bjorn Klatt, Theresia Krausl, Ullrika Sahlin, Glenn P. Svensson, Maj Rundlof
Summary: Trade-offs between pesticide use, pollinators and yield in pollinator-dependent, mass-flowering crops may cause conflicts between conservation and agronomic goals. This study proposes a framework to explore these trade-offs, using red clover as an example. The results indicate that the insecticide thiacloprid can increase seed yield without negative effects on the key pollinator, Bombus terrestris, and the presence of red clover benefits pollinator populations.
JOURNAL OF APPLIED ECOLOGY
(2022)
Article
Ecology
Maria Blasi, Yann Clough, Anna Maria Jonsson, Ullrika Sahlin
Summary: This study developed a spatially and temporally explicit theoretical model to evaluate the impact of system changes and within-season variability in resources on wild bee population sizes and crop visitation rates. The model captures the importance of food resource availability at the colony and landscape level for bee populations and crop visitation rates, as well as the potential threats of climate and land use changes to bee populations.
ECOLOGY AND EVOLUTION
(2022)
Article
Biodiversity Conservation
Georg K. S. Andersson, Niklas Boke-Olen, Fabian Roger, Johan Ekroos, Henrik G. Smith, Yann Clough
Summary: This study assesses the contributions of agricultural and forest habitats to biodiversity in boreonemoral Sweden. The results show that semi-natural pastures and cereal crops are important contributors to landscape-scale diversity, while clear-cuts provide habitats for open-land species. Maintaining farmland is critical for maintaining species richness in forestry-dominated areas.
BIOLOGICAL CONSERVATION
(2022)
Article
Biodiversity Conservation
Anna S. Persson, Amy Westman, Tobias J. Smith, Margaret M. Mayfield, Peter Olsson, Henrik G. Smith, Richard Fuller
Summary: Urbanisation can lead to declines in insect pollinators, but urban green spaces in densely populated areas can benefit a range of pollinators. The abundance of bee species was influenced by vegetation cover and human population density, with different effects on non-eusocial and eusocial species. Hoverflies were negatively affected by human density but positively influenced by vegetation cover, supporting the idea that urban greening can benefit insect pollinators.
Article
Ecology
Mike Image, Emma Gardner, Yann Clough, William E. Kunin, Simon G. Potts, Henrik G. Smith, Graham N. Stone, Duncan B. Westbury, Tom D. Breeze
Summary: This study investigates the impact of the English agri-environment scheme on pollination services and identifies hedge/woodland edge management and fallow areas as the main drivers. Floral margins have limited benefits, and interventions are more effective in areas with fewer semi-natural habitats.
Article
Biology
Oceane Bartholomee, Ciara Dwyer, Pierre Tichit, Paul Caplat, Emily Baird, Henrik G. Smith
Summary: This study revealed that bumblebee communities segregate based on light intensity, with higher investment in light sensitivity observed in darker conditions. Additionally, the eye parameter is linked to the realized niche optimum of bumblebee species. These findings suggest that microhabitat niche partitioning, based on visual traits, may contribute to the coexistence of bumblebee species.
PROCEEDINGS OF THE ROYAL SOCIETY B-BIOLOGICAL SCIENCES
(2023)
Review
Agronomy
Josefin Winberg, Henrik G. Smith, Johan Ekroos
Summary: The Paris agreement on climate change requires rapid reductions in greenhouse gas emissions, and bioenergy is seen as an important strategy in replacing fossil fuels. However, the use of biomass for energy has sparked controversy over its impact on climate mitigation and biodiversity conservation. This review provides an overview of the impacts of bioenergy crop production on ecosystems in temperate climates, highlighting the importance of factors such as land use, crop type, and scale of production in determining the effects on biodiversity and ecosystem services.
GLOBAL CHANGE BIOLOGY BIOENERGY
(2023)
Article
Biochemistry & Molecular Biology
Cecilia Kardum Hjort, Josephine R. Paris, Henrik G. Smith, Rachael Y. Dudaniec
Summary: Invasive bumblebees, such as Bombus terrestris in Tasmania, can rapidly adapt and thrive in non-native environments. This study found high gene flow but low genetic diversity in the invasive population, with restricted migration in certain regions related to elevation, land use, wind speed, and precipitation seasonality. Selection signatures were also identified for genes related to precipitation, wind speed, and wing loading. These findings provide insights into the evolutionary processes and potential global spread of invasive pollinators.
Article
Multidisciplinary Sciences
Lina Herbertsson, Bjorn K. Klatt, Maria Blasi, Maj Rundlof, Henrik G. Smith
Summary: This study investigated the impacts of neonicotinoid insecticides on bees. The results showed that bees exposed to the insecticide had slower foraging speed and reduced pollination performance, but their reproduction was not significantly affected. Although the lack of a mechanistic explanation, understanding the complex effects of plant protection products is important.
Article
Biodiversity Conservation
Klas Rydhmer, Jord Prangsma, Mikkel Brydegaard, Henrik G. Smith, Carsten Kirkeby, Inger Kappel Schmidt, Birte Boelt
Summary: This study utilized entomological lidar in a clover seed crop to profile the activity of honeybees and other insects, demonstrating the ability to record high numbers of insects in a short time period. The spatial model derived was able to effectively separate honeybees from wild insects, providing valuable insight into the distribution and activity of bees in relation to their hives and surrounding environment.
ANIMAL BIOTELEMETRY
(2022)