Output Distribution Research Articles

Mean Variance Estimation Neural Network Particle Filter for Predicting Battery Remaining Useful Life

Traditional remaining useful life (RUL) prediction methods based on particle filter (PF) require the manual tuning of hyperparameters, such as process or measurement noise, which poses challenges, particularly in real-life applications where external and operating conditions may change, potentially leading to large errors in the predictions. We address this issue by replacing the measurement equation of a PF with a mean variance estimation neural network that estimates the mean and the variance of the output distribution. As a result, the measurement noise is automatically estimated by the neural network and does not require manual setting. Through simulations and comparative analyses with state-of-the-art methods, the proposed mean variance estimation neural network particle filter (MVENN-PF) is shown to provide more stable and accurate RUL predictions, thereby potentially enhancing the robustness of battery health management systems based on it. Additionally, by eliminating the need to manually set a model hyperparameter (the measurement noise) the proposed method simplifies the modeling process, making it more accessible and adaptable to various battery systems.

Jaynes machine: The universal microstructure of deep neural networks

Despite the recent stunning progress in large-scale deep neural network applications, our understanding of their microstructure, ‘energy’ functions, and optimal design remains incomplete. Here, we present a new game-theoretic framework, called statistical teleodynamics, that reveals important insights into these key properties. The optimally robust design of such networks inherently involves computational benefit-cost trade-offs that physics-inspired models do not adequately capture. These trade-offs occur as neurons and connections compete to increase their effective utilities under resource constraints during training. In a fully trained network, this results in a state of arbitrage equilibrium, where all neurons in a given layer have the same effective utility, and all connections to a given layer have the same effective utility. The equilibrium is characterized by the emergence of two lognormal distributions of connection weights and neuronal output as the universal microstructure of large deep neural networks. We call such a network the Jaynes Machine. Our theoretical predictions are shown to be supported by empirical data from seven large-scale deep neural networks. We also show that the Hopfield network and the Boltzmann Machine are the same special case of the Jaynes Machine.

Racing Demands for Winning a Grand Tour: Differences and Similarities Between a Female and a Male Winner.

To describe and compare the race characteristics, demands, and durability profile of a male and a female Grand Tour winner. Overall and stage-type-specific (ie,time trials, flat, semimountainous, and mountain) demands and race characteristics during 2 Grand Tours were determined and compared between the female and male cyclists. In addition, relative power output distribution and pacing, percentage of functional threshold power (FTP), and changes in maximal mean power outputs (MMPs) with increasing levels of kilojoules burned were determined. Although many differences were found between course and absolute racing demands between the male (FTP: 413W; critical power: 417W) and female (FTP: 297W; critical power: 297W) cyclists, similar power distributions and pacing strategies were found if data were expressed relatively. However, the female cyclist rode a higher percentage of her FTP during the first 2 quarters of flat stages (14.7%-15.1%) and the last quarter of mountain stages (9.8%) than the male cyclist. Decrements in MMPs were only observed after burning 30kJ·kg-1 in the female and 45kJ·kg-1 in the male Grand Tour winner. Both the male and female Grand Tour winners produced very high 20- to 60-minute MMPs, whereas decrements in MMPs were only observed after having burned 75% (female) and 80% (male) of total kilojoules burned during a stage. These are the latest and lowest in MMPs reported in the scientific literature and highlight the importance of durability in combination with excellent climbing and time-trial skills, which are needed to be able to win a Grand Tour.

Mapping the Path to Low-Carbon Behaviour: A Systematic Review of Trends, Gaps, and Future Directions

It is essential to conduct research on low-carbon behaviour in order to address global climate change and promote sustainable development. This study conducts a thorough examination of a bibliometric analysis of the existing literature in this field. We analysed 129 papers from the Web of Science Core Collection database to conduct a study on the concept of “low-carbon behaviour”. Review articles, proceedings papers, and editorial materials were all excluded. This investigation examines a diverse array of research environment factors, including the most highly ranked publications, publication trends, significant publishers, and patterns in citations and publications over time. The primary institutional contributors in the discipline and the most influential works worldwide are also mentioned. The application of advanced visualisation techniques, such as wordcloud representations, the spatial distribution of research output, and co-occurrence and overlay networks, is employed to generate maps of keyword relationships and collaboration networks using Bibliometrix Stable Version, VOSviewer 1.6.20, and Scimago Graphica Beta 1.0.45. This multi-modal methodology enables a comprehensive investigation of significant research issues and emerging directions in low-carbon behavioural studies. This study contributes to the ongoing endeavours to promote sustainable development and mitigate climate change by conducting a comprehensive evaluation of the current state of research and establishing a robust framework for future investigations into low-carbon behavioural patterns and behaviours.

Simulation study of R1234ze and its mixed working medium in TEG‐ORC combined cycle

AbstractThermoelectric generation (TEG) and organic Rankine cycle (ORC) technologies both have their respective limitations in recovering waste heat from ships. However, the combination of TEG and ORC is an effective approach to achieve cascaded waste heat recovery and improve heat utilization efficiency. There has been some progress in research on waste heat recovery in maritime applications based on TEG‐ORC combined cycles. The selection of a working medium is a crucial element that directly influences the design and performance of the entire system, but there is limited research on the impact of different working fluids on combined cycle systems. In this study, a simulation model of a TEG‐ORC combined cycle system was established to examine its output potential using two different working fluids R1234ze and a mixture of R245fa/R1234ze. The results demonstrate that compared to employing the R1234ze working medium, the combined cycle system with the mixed working medium achieved a 13% increase in maximum output power, a 102% improvement in optimum thermal efficiency, and a 53% reduction in optimum power production cost. Additionally, the power output distribution in the system utilizing the mixed working fluid was more uniform compared to the system employing a single working medium. This confirms the great potential of using a mixed working fluid in a combined cycle system.

Development overview and research hotspots of Information Science & Library Science based on SSCI data

This study uses the scientometric method to analyse the literature to summarise the development overview and research hotspots in the field of Library and Information Science (LIS), aiming at sorting out the knowledge structure of the discipline and providing references and information for deepening its research and development. This article takes 85 Social Science Citation Index (SSCI) journals in the field of LIS as the research object, from which 116,570 articles from 1900 to 2023 are collected as the basis of data analysis. It employs the scientometric method based on visualisation software to analyse four dimensions of the LIS discipline: annual publications, journal outputs distribution, production and cooperation, and present focus and research hotspots, finding that the overall development of the discipline is stable and shows some new topics.

Defining a maximum heat release rate probability distribution function for design fires in sprinkler‐protected residential buildings

AbstractIn fire safety engineering analysis of sprinkler‐protected residential buildings, the maximum heat release rate is a key parameter requiring consideration. Several documents provide advice for estimating the heat release rate of a sprinkler‐controlled fire, with a prevailing suggestion that it is fixed upon activation of the first sprinkler. When carrying out deterministic analysis, this requires the engineer to assume fixed fire parameters and consider that sprinklers limit fire growth. To explore these assumptions, the study uses three deterministic models to estimate a sprinkler‐controlled maximum heat release rate for a representative apartment layout. The models include Alpert's correlation, a B‐RISK zone model and a computational fluid dynamics model in the Fire Dynamics Simulator. These deterministic models are compared to a probabilistic model in B‐RISK, where Monte Carlo simulations are used to generate a range of maximum heat release rates from distribution functions for fire and sprinkler properties. An output distribution function is generated with a mean of 296.6 kW and a standard deviation of 503.8 kW, with a lognormal distribution (μ = 5.014, σ = 1.165) estimated as a best‐fit. The deterministic models are estimated to sit in the 92–98 percentile range of this function, indicating that common deterministic assumptions are reasonably conservative. The article concludes with suggesting that, for deterministic analysis, a percentile between the 80th and 99th (340–2640 kW) could be qualitatively selected based on the design objectives, building situation and relative consequence of a fire. Further research is needed to establish guidelines for selecting appropriate percentiles across various building scenarios.

A Multiscale Mathematical Model for the Fetal Blood Circulation of the Second Half of Pregnancy.

Doppler ultrasound is a commonly used method to assess hemodynamics of the fetal cardiovascular system and to monitor the well-being of the fetus. Indices based on the velocity profile are often used for diagnosis. However, precisely linking these indices to specific underlying physiology factors is challenging. Several influences, including wave reflections, fetal growth, vessel stiffness, and resistance distal to the vessel, contribute to these indices. Understanding these data is essential for making informed clinical decisions. Mathematical models can be used to investigate the relation between velocity profiles and physiological properties. This study presents a mathematical model designed to simulate velocity wave propagation throughout the fetal cardiovascular system, facilitating the assessment of factors influencing velocity-based indices. The model combines a one-fiber model of the heart with a 1D wave propagation model describing the larger vessels of the circulatory system and a lumped parameter model for the microcirculation. Fetal growth from 20 to 40 weeks of gestational age is incorporated by adjusting cardiac and circulatory parameter settings according to scaling laws. The model's results, including cardiac function, cardiac output distribution, and volume distribution, show a good agreement with literature studies for a growing healthy fetus from 20 to 40 weeks. In addition, Doppler indices are simulated in various vessels and agree with literature as well. In conclusion, this study introduces a novel closed-loop 0D-1D mathematical model that has been verified against literature studies. This model offers a valuable platform for analyzing factors influencing velocity-based indices in the fetal cardiovascular system.

Geoinference of author affiliations using NLP-based text classification

Author affiliations are essential in bibliometric studies, requiring relevant information extraction from free-text affiliations. Precisely determining an author’s location from their affiliation is crucial for understanding research networks, collaborations, and geographic distribution. Existing geoparsing tools using regular expressions have limitations due to unstructured and ambiguous affiliations, resulting in erroneous location identification, especially for unconventional variations or misspellings. Moreover, their inefficient handling of big datasets hampers large-scale bibliometric studies. Though machine learning-based geoparsers exist, they depend on explicit location information, creating challenges when detailed geographic data is absent. To address these issues, we developed and evaluated a natural language processing model to predict the city, state, and country from an author’s free-text affiliation. Our model automates location inference, overcoming drawbacks of existing methods. Trained and tested with MapAffil, a publicly available geoparsed dataset of PubMed affiliations up to 2018, our model accurately retrieves high-resolution locations, even without explicit mentions of a city, state, or even country. Leveraging NLP techniques and the LinearSVC algorithm, our machine learning model achieves superior accuracy based on several validation datasets. This research demonstrates a practical application of text classification for inferring specific geographical locations from free-text affiliations, benefiting researchers and institutions in analyzing research output distribution.

Noise-resilient designs and analysis for optical neural networks

Abstract All analog signal processing is fundamentally subject to noise, and this is also the case in next generation implementations of Optical Neural Networks (ONNs). Therefore, we propose the first hardware-based approach to mitigate noise in ONNs. A tree-like and an accordion-like design are constructed from a given Neural Network (NN) that one wishes to implement. Both designs have the capability that the resulting ONNs give outputs close to the desired solution. To establish the latter, we analyze the designs mathematically. Specifically, we investigate a probabilistic framework for the tree-like design that establishes the correctness of the design, i.e., for any feed-forward NN with Lipschitz continuous activation functions, an ONN can be constructed that produces output arbitrarily close to the original. ONNs constructed with the tree-like design thus also inherit the universal approximation property of NNs. For the accordion-like design, we restrict the analysis to NNs with linear activation functions and characterize the ONNs’ output distribution using exact formulas. Finally, we report on numerical experiments with LeNet ONNs that give insight into the number of components required in these designs for certain accuracy gains. The results indicate that adding just a few components and/or adding them only in the first (few) layers in the manner of either design can already be expected to increase the accuracy of ONNs considerably. To illustrate the effect we point to a specific simulation of a LeNet implementation, in which adding one copy of the layers components in each layer reduces the Mean Squared Error (MSE) by 59.1% for the tree-like design and by 51.5% for the accordion-like design. In this scenario, the gap in accuracy of prediction between the noiseless NN and the ONNs reduces even more: 93.3% for the tree-like design and 80% for the accordion-like design.

Advanced tractography-guided laser ablation of a perirolandic long-term epilepsy-associated tumor: illustrative case.

Microsurgical resection of drug-resistant epilepsy-associated perirolandic lesions can lead to postoperative motor impairment. Magnetic resonance imaging (MRI)-guided laser interstitial thermal therapy (MRgLITT) has emerged as a less invasive alternative, offering reduced surgical risks and improved neurological outcomes. Electrophysiological tools routinely used for motor mapping in resective microsurgery are incompatible with intraoperative MRI. The utilization of advanced neuroimaging adjuncts for eloquent brain mapping during MRgLITT is imperative. The authors present the case of a 17-year-old athlete who underwent MRgLITT for a perirolandic long-term epilepsy-associated tumor (LEAT). They performed probabilistic multi-tissue constrained spherical deconvolution (MT-CSD) tractography to delineate the corticospinal tract (CST) for presurgical planning and intraoperative image guidance. The CST tractography was integrated into neuronavigation and MRgLITT workstation software to guide the ablation while monitoring the CST throughout the procedure. The integration of CST tractography into neuronavigation workstation planning and laser ablation workstation thermoablation is feasible and practical, facilitating complete ablation of a deep-seated perirolandic LEAT while preserving motor function. Probabilistic MT-CSD tractography enhanced MRgLITT planning as well as intraprocedural CST visualization and preservation, leading to a favorable functional outcome. The limitations of tractography and the predictability of thermal output distribution compared to the gold standard of microsurgical resection merit further discussion. https://thejns.org/doi/10.3171/CASE24139.

EMPIRICAL ESTIMATION OF THE POTENTIAL ECONOMIC IMPACT OF TOURISM PROMOTION POLICY IN ALGERIA

Objective: The objective of this paper is to empirically assess the economic impact of tourism promotion policies in Algeria. By utilizing an economic model based on the 2021 Social Accounting Matrix (SAM) multipliers, the study aims to quantify the direct, indirect, and induced effects of increased tourism demand on various sectors of the national economy. Specifically, the research seeks to evaluate how tourism expenditure influences production, sectoral output, and income distribution, while also identifying which sectors and economic agents benefit the most from tourism growth. Theoretical Framework: Study is based on the Social Accounting Matrix (SAM) approach, which examines the interconnections between sectors of the economy and tracks the flow of income and expenditure. It incorporates input-output theory to measure the direct, indirect, and induced effects of increased tourism demand. The concept of economic multipliers is central, highlighting how tourism spending influences various sectors. Additionally, the framework assesses the distribution of benefits across different economic agents, focusing on how tourism impacts production, sectoral output, and income distribution in Algeria. Method: This study uses 2021 Social Accounting Matrix (SAM) data to analyze tourism demand in Algeria. It develops an economic model, simulates scenarios, and assesses the economic gains and distribution of benefits, providing a comprehensive evaluation. Results and Discussion: The SAM-based economic model shows that Algeria's tourism demand boosts production and output across sectors, particularly agriculture, hospitality, agri-food, construction, transportation, and energy. However, the benefits are not evenly distributed, potentially leading to income inequality. Policy interventions should focus on equitable income distribution and supporting key industries. Research Implications: Tourism can drive Algeria's economic diversification, reducing dependence on traditional sectors. Inclusive tourism policies are needed to ensure equitable benefits for wage earners and lower-income groups. Future research should focus on sector investment and employment effects. Originality/Value: This study analyzes the economic impacts of tourism promotion policies in Algeria, focusing on the 2021 Social Accounting Matrix. It provides insights into the distribution of benefits and emphasizes the need for inclusive policies. The findings provide actionable insights for policy development.

Agri-Food Sector in Ukraine and Poland: A Comparative Analysis Using the Input–Output Model

The war in Ukraine and the COVID-19 pandemic have shown that Ukraine is strongly integrated with global agricultural markets, is of great importance for the world’s food security, and can effectively compete with Poland in exporting to EU markets. Contrary to Poland, the agri-food sector in Ukraine has not yet been the subject of research using input–output tables. The input–output model has many advantages in studying the structures of individual sectors and their connections with other sectors of the economy and with foreign countries. Therefore, the current study fills the existing research gap. The aim of the article is a comparative analysis of the agri-food sectors in Ukraine and Poland in 2020 (during the COVID-19 pandemic) using the input–output model. The data source is the OECD TiVA database. The main conclusion is that Poland is further along the path of agricultural and food sector development than Ukraine. The study also reveals that in terms of value, the agri-food sector in Ukraine is significantly smaller than in Poland, but it is more important in the Ukrainian economy, with agriculture being its dominant link, while in Poland it is the food industry. The Polish agri-food sector is characterized by a higher import dependency of gross output. Exports play a significant role in the distribution of the gross output of Ukraine’s agriculture and Poland’s food industry. However, the export directions differed significantly. Poland exported mainly to the EU, while Ukraine exported to Asian and Middle Eastern countries.

Improving stability and performance of spiking neural networks through enhancing temporal consistency

Spiking neural networks have gained significant attention due to their brain-like information processing capabilities. The use of surrogate gradients has made it possible to train spiking neural networks with backpropagation, leading to impressive performance in various tasks. However, spiking neural networks trained with backpropagation typically approximate actual labels using the average output, often necessitating a larger simulation timestep to enhance the network’s performance. This delay constraint poses a challenge to the further advancement of spiking neural networks. Current training algorithms tend to overlook the differences in output distribution at various timesteps. Particularly for neuromorphic datasets, inputs at different timesteps can cause inconsistencies in output distribution, leading to a significant deviation from the optimal direction when combining optimization directions from different moments. To tackle this issue, we have designed a method to enhance the temporal consistency of outputs at different timesteps. We have conducted experiments on static datasets such as CIFAR10, CIFAR100, and ImageNet. The results demonstrate that our algorithm can achieve comparable performance to other optimal SNN algorithms. Notably, our algorithm has achieved state-of-the-art performance on neuromorphic datasets DVS-CIFAR10 and N-Caltech101, and can achieve superior performance in the test phase with timestep T=1.

Training for Elite Team-Pursuit Track Cyclists-Part I: A Profile of General Training Characteristics.

To profile the training characteristics of an elite team pursuit cycling squad and assess variations in training intensity and load accumulation across the 36-week period prior to a world-record performance at the 2018 Commonwealth Games. Training data of 5 male track endurance cyclists (mean [SD]; age 21.9 [3.52]y; 4.4 [0.16]W·kg-1 at anaerobic threshold; 6.2 [0.28]W·kg-1 maximal oxygen uptake 68.7 [2.99]mL kg·min-1) were analyzed with weekly total training volume and heart rate, power output, and torque intensity distributions calculated with reference to their 3:49.804min:s.ms performance requirements for a 4-km team pursuit. Athletes completed 543 (37)h-1 of training across 436 (16) sessions. On-bike activities accounted for 69.9% of all training sessions, with participants cycling 11,246 (1139)km-1 in the training period of interest, whereas 12.7% of sessions involved gym/strength training. A pyramidal intensity distribution was evident with over 65% and 70% of training, respectively, performed at low-intensity zone heart rate and power output, whereas 5.3% and 7.7% of training was performed above anaerobic threshold. The athletes accumulated 4.4% of total training volume at, or above, their world-record team pursuit lead position torque (55N·m). These data provide updated and novel insight to the power and torque demands and load accumulation contributing to world-record team pursuit performance. Although the observed pyramidal intensity distribution is common in endurance sports, the lack of shift toward a polarized intensity distribution during taper and competition peaking differs from previous research.

Training for Elite Team-Pursuit Track Cyclists-Part II: A Comparison of Preparation Phases in Consecutive World-Record-Breaking Seasons.

To compare the training characteristics of an elite team pursuit cycling squad in the 3-month preparation phases prior to 2 successive world-record (WR) performances. Training data of 5 male track endurance cyclists (mean [SD]; age 23.4 [3.46]y; body mass 80.2 [2.74]kg; 4.5 [0.17]W·kg-1 at LT2; maximal aerobic power6.2 [0.27]W·kg-1; maximal oxygen uptake 65.9 [2.89]mL·kg-1·min-1) were analyzed with weekly total training volume by training type and heart rate, power output, and torque intensity distributions calculated with reference to the respective WRs' performance requirements. Athletes completed 805 (82.81) and 725 (68.40)min·wk-1 of training, respectively, in each season. In the second season, there was a 32% increase in total track volume, although track sessions were shorter (ie,greater frequency) in the second season. A pyramidal intensity distribution was consistent across both seasons, with 81% of training, on average, performed below LT1 power output each week, whereas 6% of training was performed above LT2. Athletes accumulated greater volume above WR team pursuit lead power (2.4% vs 0.9%) and torque (6.2% vs 3.2%) in 2019. In one athlete, mean single-leg-press peak rate of force development was 71% and 46% higher at mid- and late-phases, respectively, during the preparation period. These findings provide novel insights into the common and contrasting methods contributing to successive WR team pursuit performances. Greater accumulation of volume above race-specific power and torque (eg,team pursuit lead), as well as improved neuromuscular force-generating capacities, may be worthy of investigation for implementation in training programs.

Predicting the heat release variability of Li-ion cells under thermal runaway with few or no calorimetry data

Accurate measurement of the variability of thermal runaway behavior of lithium-ion cells is critical for designing safe battery systems. However, experimentally determining such variability is challenging, expensive, and time-consuming. Here, we utilize a transfer learning approach to accurately estimate the variability of heat output during thermal runaway using only ejected mass measurements and cell metadata, leveraging 139 calorimetry measurements on commercial lithium-ion cells available from the open-access Battery Failure Databank. We show that the distribution of heat output, including outliers, can be predicted accurately and with high confidence for new cell types using just 0 to 5 calorimetry measurements by leveraging behaviors learned from the Battery Failure Databank. Fractional heat ejection from the positive vent, cell body, and negative vent are also accurately predicted. We demonstrate that by using low cost and fast measurements, we can predict the variability in thermal behaviors of cells, thus accelerating critical safety characterization efforts.

Determination of neutron spectrum based on artificial neural network using liquid scintillation detector EJ-301.

This paper focuses on the neutron spectrum measurement using a liquid scintillation detector, where the neutron spectrum could be identified and unfolded from the light output distribution of the EJ-301 liquid scintillation detector through a linear artificial neural network (ANN). The response functions of the EJ-301 detector for monoenergetic neutron sources, as well as the light outputs, have been simulated and calculated by Monte Carlo procedure FLUKA. The linear ANN was trained and tested through the simulated data, where response functions were set as the input of ANN and the corresponding neutron spectra were output. Therefore, the neutron spectrum-unfolding model was created. This spectrum-unfolding model was tested through the light outputs induced by monoenergetic neutrons and the random superposition of them. Unfolding results show that this model could identify the information of the neutron spectrum accurately from the light outputs of a liquid scintillation detector. Moreover, the EJ-301 detector was used to measure the radioactivity of 252Cf, and the pulse height distribution induced by neutrons was derived through the charge-comparison method to remove the influence of gamma rays. The measured pulse height distribution was unfolded by the trained model, and measured results show that the unfolded neutron spectrum of 252Cf was consistent with the reference one. This paper presents the feasibility that the unknown neutron spectrum could be identified and confirmed through a linear neural network trained by simulated monoenergetic neutron response functions, which could be a candidate of choice for the determination of the neutron spectrum.

Convergence analysis of a novel high order networks model based on entropy error function

It is generally known that the error function is one of the key factors that determine the convergence, stability and generalization ability of neural networks. For most feedforward neural networks, the squared error function is usually chosen as the error function to train the network. However, networks based on the squared error function can lead to slow convergence and easily fall into local optimum in the actual training process. Recent studies have found that, compared to the squared error function, the gradient method based on the entropy error function measures the difference between the probability distribution of the model output and the probability distribution of the true labels during the iterative process, which can be more able to handle the uncertainty in the classification problem, less likely to fall into a local optimum and can learn to converge more rapidly. In this paper, we propose a batch gradient method for Sigma-Pi-Sigma neural networks based on the entropy error function and rigorously demonstrate the weak and strong convergence of the new algorithm in the batch input mode. Finally, the theoretical results and effectiveness of the algorithm are verified by simulation.

Making hospitals innovative: Macro-level policy to sustain micro-innovations in healthcare.

Successful innovation clusters are notoriously difficult to establish, and many attempts fail. How can we go about designing such systems reliably? We describe how ecosystems can be strengthened through grassroots bottom-up efforts that empower user and community innovation, as opposed to economic policies that dictate innovation. Specifically focusing on the healthcare industry, we advocate that community hospitals which constitute 90% of all hospitals in Canada are the ideal setting for such community innovation efforts. We investigated the distribution of innovation output from hospitals over the past 13years and found a decrease in predominance of major teaching hospitals, supporting the potential role for community hospitals in this space. We categorize different types of innovations and recommend institutional policies that can sustain bottom-up, micro-level efforts. Such policies could improve and enhance the development of micro-innovations and the creation of health innovation clusters.