Flow-based Model Research Articles

A Microscopic On-Ramp Model Based on Macroscopic Network Flows

While macroscopic traffic flow models adopt a fluid dynamic description of traffic, microscopic traffic flow models describe the dynamics of individual vehicles. Capturing macroscopic traffic phenomena accurately remains a challenge for microscopic models, especially in complex road sections. Based on a macroscopic network flow model calibrated to real traffic data and new rules for the acceleration and merging behavior on the on-ramp, we propose a microscopic model for on-ramps. To evaluate the performance of the new flow-based model, we conduct traffic simulations assessing speeds, accelerations, lane change positions, and risky behavior. Our results show that, although the proposed model exhibits some limitations, its performance is superior to the Intelligent Driver Model in the evaluated aspects. While the Intelligent Driver Model simulations are almost free of conflicts, the proposed model evokes a realistic amount and severity of conflicts and therefore can be considered for safety analysis.

A comparative lattice analysis of SU(2) dark glueballs* *Supported in part by the Natural Science Foundation of China (12125503, 12335003), and the Natural Science Foundation of Shandong Province, China (ZR2022ZD26)

We study the mass and scattering cross section of glueballs as dark matter candidates using lattice simulations. We employ both naive and improved gauge actions in dimensions with several β values, and we adopt both the traditional Monte Carlo method and flow-based model based on machine learning techniques to generate lattice configurations. The mass of a dark scalar glueball with and the Nambu-Bethe-Salpeter wave function are calculated. Using a coupling constant of as an illustration, we compare the dark glueball mass calculated from the configurations generated from the two methods. While consistent results can be achieved, the two methods demonstrate distinct advantages. Using the Runge-Kutta method, we extract the glueball interaction potential and two-body scattering cross section. From the observational constraints, we obtain the lower bound of the mass of scalar glueballs as candidates of dark matter.

The relevance of cash flow information in predicting corporate bankruptcy in Italian private companies

This study investigates the relevance of cash flow information in predicting corpo-rate bankruptcy in Italian private companies. The results indicate that while the cash flow-based model exhibits a high predictive capacity, it is less effective than the accrual-based model. In addition, cash flow-based ratios do not improve the predictive capacity of the accrual-based model. From a theoretical perspective, this study enriches existing literature on the relevance of cash flow information in predicting corporate bankruptcy by extending the investigation to the Italian con-text, which has not yet been sufficiently studied. From a practical standpoint, it provides Italian companies with new bankruptcy prediction models and offers pre-liminary suggestions regarding the relevance that should be attributed to cash flow information within the organizational, administrative, and accounting structures that they must establish to promptly detect crises and undertake appropriate initia-tives in a timely manner to comply with the requirements of the new legislation on business crises.

Collaborative optimization of depot location, capacity and rolling stock scheduling considering maintenance requirements

Generally, when optimizing a rolling stock schedule, the locations of the depots, or places in the network where the composition changes and maintenance occurs, are assumed known. The locations where maintenance is performed naturally influence the quality of any resulting rolling stock schedules. In this paper, the problem of selecting new depot locations and their corresponding capacities is considered. A two-stage mixed integer programming approach for rolling stock scheduling with maintenance requirements is extended to account for depot selection. First, a conventional flow-based model is solved, ignoring maintenance requirements, to obtain a variety of rolling stock schedules with multiple depot locations and capacity options. Then, a maintenance feasible rolling stock schedule can be obtained by solving a series of assignment problems by using the schedules found in the first stage. The proposed methodology is tested on real-life instances, and the numerical experiments of different operational scenarios are discussed.

FeatsFlow: Traceable representation learning based on normalizing flows

This paper studies effective traceable feature representation learning in the view of distribution transformation, termed FeatsFlow, by proposing a distribution-aware learning framework combining the discriminating model with a normalizing flow-based model. The process can be regarded as a series of feature distribution transformations, from the input images to the expected results. Focusing on the learned representation of the target model, we take full advantage of the invertible nature of normalizing flows and learn the practical and traceable feature representation for target goals. Considering that it is difficult to model the traceable process for feature extraction, we propose an effective model by combining a general discriminating model with normalizing flows for traceable feature extraction. The normalizing flows module is added to the original model in a plug-in mode, which is convenient to make it available for effective and traceable feature learning. Thus we can obtain an effective and traceable representation distribution. Extensive experiments are conducted on our proposed representation learning model for the image classification task, and the experimental results illustrate that our proposed model is adequate for traceable representation learning. The most important is that we present a distribution-aware representation learning approach, which makes it possible to conduct and understand feature representation learning at the feature level.

Coordinated flow model for strategic planning of autonomous mobility-on-demand systems

ABSTRACT High-quality strategic planning of autonomous mobility-on-demand (AMOD) systems is critical for the success of the subsequent phases of AMOD system implementation. To assist in strategic AMOD planning, we propose a dynamic and flexible flow-based model of an AMOD system. The proposed model is computationally fast while capturing the state transitions of two coordinated flows (i.e. co-flows): the AMOD service fleet vehicles and AMOD customers. Capturing important quantity dynamics and conservations through a system of ordinary differential equations, the model can economically respond to a large number and a wide range of scenario-testing requests. The paper illustrates the model efficacy through a basic example and a more realistic case study. The case study envisions replacing Manhattan's existing taxi service with a hypothetical AMOD system. The results show that even a simple co-flow model can robustly predict the systemwide AMOD dynamics and support the strategic planning of AMOD systems.

Semi-Supervised Deep Learning-Based Multi-component Spectral Calibration Modeling for UV-vis and Near-Infrared Spectroscopy without Information Loss.

Spectral analysis is an important method for characterizing and identifying chemical species. However, quantitative spectral analysis of multiple chemical properties in the real world has always been a challenging problem due to the strong correlation, massive noise, and serious information overlapping of the spectral features. Here, we present a new semi-supervised spectral calibration method based on information lossless decoupling of spectral features named NICEM. To realize the separation and extraction of key latent features, the method uses the flow-based model non-linear independent component estimation (NICE) to learn the sample distribution. The spectral data information is transformed into independent latent variables obeying Gaussian distribution by the reversible structure of deep network without information loss, so as to find the essential properties and realize the feature nonlinear decomposition. Moreover, the association between the input latent feature variables and attributes is evaluated by the maximum mutual information coefficient to eliminate the adverse effects of irrelevant information in the latent variable space and mine key information. Since the latent variables are independent in each dimension, the NICEM method is easier to establish an accurate semi-supervised multi-component calibration model even for high overlapping and complex spectral data. The applicability of the proposed spectral modeling method is demonstrated by using three ultraviolet-visible and near-infrared spectral data sets with 15 physical and chemical properties including diesel fuels, corn, and multi-metal ions solution. Results show that the proposed NICEM method has the highest determination coefficient (R2) and significantly improves extrapolation compared with the seven state-of-the-art methods. The proposed method is intuitive because it obviates complex feature engineering and prior knowledge and is a promising spectral calibration tool for quantitative analysis in other spectroscopy applications.

Study of reliable routing process to ensure the communication network resilience

The article is devoted to a relevant scientific and applied problem related to improving potential solutions to increase network resilience by means of reliable routing. The study of several routing models is presented, taking into account different types of metrics, in particular, the metric based on the availability indicator. The technological task of routing is formulated in an optimization form. Attention is focused on the flow-based model of reliable routing, and its effectiveness is proved through modeling and comparative analysis with existing approaches. Also, single-path and multipath routing strategies were analyzed in terms of their effectiveness in increasing the level of network resilience. It is demonstrated that the multipath approach to reliable routing is necessary at high loads and contributes to an increase in the overall multi-path availability. At the same time, the implementation of a single-path routing strategy at high loads leads to routing solutions with a low availability indicator and limited use of the available network resource – bandwidth. The numerical research results have confirmed the effectiveness of the proposed multipath routing solution with the availability metric and allowed the formulation of general recommendations related to the implementation of reliable routing in practice. Thus, it is proposed to use a multipath routing strategy to balance the load in the network, together with consideration of availability metrics in calculating multipath for the data flow transmission. Further improvement of the reliable routing model is envisaged, considering the Quality of Service indicators when calculating reliable and productive multipath.

Research of a flow-based model of secure Traffic Engineering routing in an infocommunication network with normalized link-blocking conditions

The work improves the flow-based model of secure routing with load balancing in accordance with the Traffic Engineering concept by considering the network security parameters in an information and communication network (ICN). Within the framework of the proposed model, the solution to the technological problem of secure routing with load balancing in an ICN was reduced to solving an optimization problem of linear programming, which guarantees the predictable computational complexity of solutions and low requirements for the computing power of devices responsible for solving routing problems – routers, route servers, controllers, etc. The novelty of the proposed model is the modification of the exponential model of blocking communication links, taking into account normalized conditions to prevent a situation where even the most secure links will be blocked, leading to inefficient use of the link resource. The results of the study of secure routing processes with load balancing in ICN confirmed the model’s effectiveness in considering the network state: its topology, flow characteristics, bandwidth, and congestion of communication links, as well as the probability of their compromise. This made it possible to orient the resulting routing solutions to reduce the congestion of communication links that have a high compromise probability by redistributing traffic to more secure links. In the course of the study, a comparative analysis of the effectiveness of using secure and TE routing models was carried out by a number of indicators. It has been established that the use of SecTE and NormSecTE secure routing models focused on finding a compromise between the Quality of Service and network security indicators. At the same time, the proposed improved model of secure TE routing NormSecTE, based on a more accurate consideration of the probability of compromising communication links, allowed, compared to the SecTE model, to improve the level of network security (packet compromise probability), but with a certain decrease in the level of Quality of Service (average end-to-end packet delay) in the ICN.

Exploring the tradeoff between energy dissipation, delay, and the number of backbones for broadcasting in wireless sensor networks through goal programming

Broadcasting, which is an essential mode of operation in wireless sensor networks (WSNs), dissipates a non-negligible portion of the energy budget of a sensor node. Broadcasting is achieved by the dissemination of broadcast packets (originated at the BS) by a set of relay nodes, which constitute a backbone so that all sensor nodes receive broadcast packets. Utilization of multiple backbones is necessary to achieve balanced energy dissipation of sensor nodes in broadcasting. In this study, we propose two mixed integer programming (MIP) models (i.e., the flow-based model and the node-based model), which minimize the energy dissipation of the highest energy-consuming node for broadcasting by utilizing multiple backbones. Balanced energy minimization and delay minimization objectives are integrated through a goal programming (GP) framework built upon the foundation provided by the scalable node-based model. Performance evaluations based on the optimal solutions of our models reveal that maximum energy dissipation and delay in WSN broadcasting can be significantly reduced, simultaneously, by utilizing multiple backbones (e.g., with two backbones maximum energy dissipation and delay can be, concurrently, reduced by more than 2% and 22%, respectively, in comparison to the single backbone case, likewise, it is also possible to simultaneously reduce maximum energy dissipation and delay more than 8% and 13%, respectively, depending on the priorities assigned to the objectives). Nevertheless, employing more than two backbones does not provide any significant performance improvements.

Flow-enhanced transportation for anomaly detection

Resonant anomaly detection is a promising framework for model-independent searches for new particles. Weakly supervised resonant anomaly detection methods compare data with a potential signal against a template of the Standard Model (SM) background inferred from sideband regions. We propose a means to generate this background template that uses a flow-based model to create a mapping between high-fidelity SM simulations and the data. The flow is trained in sideband regions with the signal region blinded, and the flow is conditioned on the resonant feature (mass) such that it can be interpolated into the signal region. To illustrate this approach, we use simulated collisions from the Large Hadron Collider (LHC) Olympics Dataset. We find that our flow-constructed background method has competitive sensitivity with other recent proposals and can therefore provide complementary information to improve future searches.

Entropy Generation Minimization for Boiling Flow Inside Evaporator Tube With R32 and R410A Refrigerants: A Comparison of Different Two-Phase Flow Models

AbstractThe objective of the present work is to find out the optimal mass flux of two common refrigerants, R32 and R410A, undergoing boiling in an evaporator tube based on the entropy generation minimization approach. An entropy generation model is developed for two-phase boiling flow based on the drift flux approach. The distinct terms for entropy generation due to heat transfer and pressure drop are developed. The optimal solutions of the present model under a practical range of parametric conditions are compared with that obtained from the existing homogeneous and separated flow-based model. Analysis reveals that increasing the evaporator diameter and heat flux increases the optimal mass flux at minimum entropy generation for both refrigerants. On the other hand, increasing the evaporator length decreases the optimal mass flux. It is observed that the optimal mass flux data for all parametric conditions deviate within 10–15% on the positive and negative sides for drift flux and homogeneous models, respectively, compared to separate flow model. The present analysis also reveals that the overall deviation of data of the drift flux model and homogeneous flow model indicates a standard deviation of 1.95–1.98%, respectively.

DualFlow: Generating imperceptible adversarial examples by flow field and normalize flow-based model.

Recent adversarial attack research reveals the vulnerability of learning-based deep learning models (DNN) against well-designed perturbations. However, most existing attack methods have inherent limitations in image quality as they rely on a relatively loose noise budget, i.e., limit the perturbations by L p -norm. Resulting that the perturbations generated by these methods can be easily detected by defense mechanisms and are easily perceptible to the human visual system (HVS). To circumvent the former problem, we propose a novel framework, called DualFlow, to craft adversarial examples by disturbing the image's latent representations with spatial transform techniques. In this way, we are able to fool classifiers with human imperceptible adversarial examples and step forward in exploring the existing DNN's fragility. For imperceptibility, we introduce the flow-based model and spatial transform strategy to ensure the calculated adversarial examples are perceptually distinguishable from the original clean images. Extensive experiments on three computer vision benchmark datasets (CIFAR-10, CIFAR-100 and ImageNet) indicate that our method can yield superior attack performance in most situations. Additionally, the visualization results and quantitative performance (in terms of six different metrics) show that the proposed method can generate more imperceptible adversarial examples than the existing imperceptible attack methods.

The Snake Eggs Puzzle: Preparing Students for Benders Decomposition

Logic puzzles are an effective way to introduce students to advanced solution techniques in operations research, such as Lagrangian relaxation, Dantzig-Wolfe decomposition, and Benders decomposition. The Snake Egg puzzle asks the player to draw a one-cell wide path, or “snake,” in a grid. The remaining cells should form a fixed number of separate, connected, discontiguous regions called “eggs.” We propose two solution approaches: a flow-based model and lazy constraints. Instead of providing the complete model at the outset, we will step through the puzzle in a manner suitable to the classroom, emphasizing the skills that are crucial to successfully implementing advanced techniques. The puzzle functions in particular as a prelude to Benders decomposition. Funding: M. Harris is supported by an Australian Government RTP (research training program) scholarship.

Explore the Feeling of Presence and Purchase Intention in Livestream Shopping: A Flow-Based Model

Livestream shopping has attracted great attention in an increasingly digitalized society. This study is to explore the mechanism through which social presence and physical presence affect consumer purchase intentions in livestream shopping as an emerging e-commerce model. Based on the flow theory, this study proposes an integrated model to explain the mechanism through which the feeling of presence affects consumers’ purchase intentions in livestream shopping. Empirical data on livestream shopping were collected in China to test the proposed model for an exploratory study. The results show that the feeling of physical presence influences consumers’ purchase intentions through concentration and perceived control, and the feeling of social presence influences consumers’ purchase intentions through concentration and enjoyment, and, thus, both social presence and physical presence are important elements in livestream shopping. This study provides a better understanding on the mechanism of how the feeling of presence helps improve purchase intentions in livestream shopping. This study shows both physical presence and social presence are positively related to consumers’ purchase intention, but with different paths, and, thus, sheds new lights on the feeling of presence and its impact on consumer behaviors in e-commerce.

A NEW FLOW-BASED MODEL OF END-TO-END INTEGRATION IN PREMIUM PRODUCT SUPPLY CHAINS

Driven by the total cost of ownership, US–China trade and technology competition, and the COVID-19 pandemic, global supply chains are undergoing a major restructuring that will soon transform business and economics all over the world. Recently, supply chains with end-to-end integration for premium agri-food products have emerged as a new business model. This paper examines how they should function, and identifies the supply chain structure/production/business conditions necessary for their development. We study a premium supply chain consisting of many small farms that produce top-quality bananas, one integrator firm, and thousands of retail stores. We use industry and business data to calibrate a multiple-route flow-based model from farms to integrator to retailers/markets. We then use sensitivity analysis to illuminate the stakeholders’ decision behaviour, and identify and discuss three main decision issues: contract farming, capacity strategy, and business robustness. For contract specification, contracting on price rather than quantity is conducive to coordinating the interests of the stakeholders. For the capacity strategy, the integrator should source raw products from many small farms rather than fewer large farms. For business robustness, the integrator could still ensure robust profits by regulating its product supply when new competitors arise or demand changes. These results are tested under various scenarios to determine the impact of input parameters or supply chain structure, and are verified with an industry practitioner who has experience with multiple premium agri-food products. The results, along with the flow model and its computation procedure, could be used by supply chain planners to start new businesses or to differentiate retailers’ premium product offerings in competitive business environments.

An imbalanced binary classification method via space mapping using normalizing flows with class discrepancy constraints

For imbalanced binary classification, most methods classify data in the original space by balancing data or modifying classification algorithms. However, for datasets with severe overlaps and complex class boundaries, the classification performance is limited by the ability of the classifier to fit the decision boundary. A practical alternative is to map the data to a latent space with lower classification difficulty. In this paper, we propose an imbalanced binary classification method via space mapping using normalizing flows with class discrepancy constraints. The flow-based model is employed to map the original data to a latent space, ensuring the accuracy of the mapping and the authenticity of the mapped distribution. To solve the problem of overlaps and within-class imbalance, class discrepancy constraints, including global and local constraints, are proposed to modify the flow-based model. The former maps different subclusters to the same cluster of the latent space, and the latter increases the separability of different classes. Classification thus can be carried out in the latent space with simpler distribution and high separability to achieve better classification performance. Experimental results on 35 KEEL and UCI public datasets indicate that the proposed method outperforms other state-of-the-art methods on F1-measure and G-mean.

Learning low-dose CT degradation from unpaired data with flow-based model.

There has been growing interest in low-dose computed tomography (LDCT) for reducing the X-ray radiation to patients. However, LDCT always suffers from complex noise in reconstructed images. Although deep learning-based methods have shown their strong performance in LDCT denoising, most of them require a large number of paired training data of normal-dose CT (NDCT) images and LDCT images, which are hard to acquire in the clinic. Lack of paired training data significantly undermines the practicability of supervised deep learning-based methods. To alleviate this problem, unsupervised or weakly supervised deep learning-based methods arerequired. We aimed to propose a method that achieves LDCT denoising without training pairs. Specifically, we first trained a neural network in a weakly supervised manner to simulate LDCT images from NDCT images. Then, simulated training pairs could be used for supervised deep denoisingnetworks. We proposed a weakly supervised method to learn the degradation of LDCT from unpaired LDCT and NDCT images. Concretely, LDCT and normal-dose images were fed into one shared flow-based model and projected to the latent space. Then, the degradation between low-dose and normal-dose images was modeled in the latent space. Finally, the model was trained by minimizing the negative log-likelihood loss with no requirement of paired training data. After training, an NDCT image can be input to the trained flow-based model to generate the corresponding LDCT image. The simulated image pairs of NDCT and LDCT can be further used to train supervised denoising neural networks fortest. Our method achieved much better performance on LDCT image simulation compared with the most widely used image-to-image translation method, CycleGAN, according to the radial noise power spectrum. The simulated image pairs could be used for any supervised LDCT denoising neural networks. We validated the effectiveness of our generated image pairs on a classic convolutional neural network, REDCNN, and a novel transformer-based model, TransCT. Our method achieved mean peak signal-to-noise ratio (PSNR) of 24.43dB, mean structural similarity (SSIM) of 0.785 on an abdomen CT dataset, mean PSNR of 33.88dB, mean SSIM of 0.797 on a chest CT dataset, which outperformed several traditional CT denoising methods, the same network trained by CycleGAN-generated data, and a novel transfer learning method. Besides, our method was on par with the supervised networks in terms of visualeffects. We proposed a flow-based method to learn LDCT degradation from only unpaired training data. It achieved impressive performance on LDCT synthesis. Next, we could train neural networks with the generated paired data for LDCT denoising. The denoising results are better than traditional and weakly supervised methods, comparable to supervised deep learningmethods.

현금흐름 기반 동태적 주택시장 모형의 구조와 시사점

It is not appropriate to apply the demand/supply approach to the housing market and it causes many problems. Unlike the premise of demand/supply approach, transactions of existing houses account for the majority of the domestic housing market, and transactions of houses with 'jeonse' are also significant. In this paper, an analysis based on cash flow is proposed as an alternative housing market analysis method. According to this, the house price is determined at the point where the amount of cash payment and the amount of cash receipt match for the sale in the period, and changes when a difference occurs between the amounts. As a result of constructing and simulating a dynamic model for housing market based on cash flow, it was estimated that more than half of the rise in Korean housing prices since 2000 was due to the rise in 'jeonse' prices, and the rest were due to the higher ratio of additional procurement cash than the equilibrium value. According to the cash flow-based analysis, heavy comprehensive real estate tax and transfer tax for multiple homeowners, house supply through the development of new cities and the unitary deregulation of reconstruction can also lead to a rise in housing prices, contrary to policy intentions. The cash flow-based dynamic market model presented in this paper can be expanded throughout the resell market where used goods with scarcity are traded

The interaction between anti-PF4 antibodies and anticoagulants in vaccine-induced thrombotic thrombocytopenia

AbstractLife-threatening thrombotic events at unusual sites have been reported after vector-based vaccinations against severe acute respiratory syndrome coronavirus 2. This phenomenon is now termed vaccine-induced immune thrombotic thrombocytopenia (VITT). The pathophysiology of VITT is similar to that of heparin-induced thrombocytopenia (HIT) and is associated with platelet-activating antibodies (Abs) against platelet factor 4 (PF4). Therefore, current guidelines suggest nonheparin anticoagulants to treat VITT patients. In this study, we investigated the interactions of heparin, danaparoid, fondaparinux, and argatroban with VITT–Ab/PF4 complexes using an ex vivo model for thrombus formation as well as in vitro assays to analyze Ab binding and platelet activation. We found that immunoglobulin Gs (IgGs) from VITT patients induce increased adherent platelets/thrombus formation in comparison with IgGs from healthy controls. In this ex vivo flow-based model, the procoagulant activity of VITT IgGs was effectively inhibited with danaparoid and argatroban but also by heparin. Interestingly, heparin and danaparoid not only inhibited IgG binding to PF4 but were also able to effectively dissociate the preformed PF4/IgG complexes. Fondaparinux reduced the in vitro generation of procoagulant platelets and thrombus formation; however, it did not affect platelet aggregation. In contrast, argatroban showed no effect on procoagulant platelets and aggregation but significantly inhibited VITT-mediated thrombus formation. Taken together, our data indicate that negatively charged anticoagulants can disrupt VITT–Ab/PF4 interactions, which might serve as an approach to reduce Ab-mediated complications in VITT. Our results should be confirmed, however, in a clinical setting before a recommendation regarding the selection of anticoagulants in VITT patients could be made.