Search

The recognition of medical images, especially endoscopic ultrasound images, has the characteristics of changing images and insignificant gray-scale changes, which requires repeated observation and comparison by medical staff. In view of the above-mentioned characteristics of ultrasound imaging, a system scheme suitable for image processing is proposed, which can analyze the biliary tract, gallbladder, abdominal lymph nodes, liver, descending duodenum, duodenal bulb, stomach, pancreas, pancreatic lymph nodes, there are a total of 10 ultrasonic organs, including 21 kinds of sub-categories and 3510 images. The images are preprocessed using binarization, histogram equalization, median filtering and edge enhancement algorithms. The improved YoloV4 convolutional neural network algorithm is used to train the data set and perform high accuracy is detected in real time. Finally, the average accuracy of this algorithm has reached 91.59%. The algorithm proposed in this paper can make up for the shortcomings of manual detection in the original image detection system, improve the efficiency of detection, and at the same time as an auxiliary system can reduce detection misjudgments, and promote the development of automated and intelligent detection in the medical field.

Distributed Generations are usually connected to the ac grid through HVDC transmission systems. However, due to the current injected by the grid-side voltage source converter (GVSC), the voltage drop after a short-circuit fault will be partly compensated. For this reason, the viewed impedance of distance relays will be affected, which is a challenge for the distance protection scheme. For this issue, this paper proposed an adaptive distance protection scheme for the system containing distributed generations. Firstly, the proposed scheme controls the current of the GSVC to be proportional to the current of the AC transmission line. Secondly, an adaptive distance protection criterion is proposed, which can obtain the impedance from the relay to the fault-point accurately. The effectiveness of the proposed approach was validated through simulation using MATLAB/Simulink.

Microwave diplexers are often designed iteratively via manual channel-by-channel or resonator-by-resonator local optimization procedures due to high dimensionality in design variables and the complexity of structures. A one-off automated diplexer optimization method is yet to be developed. To achieve this, other than improved optimizers, an effective objective function (OF) is critical to allow optimizers to find the desired designs. This letter realizes an automated diplexer design by proposing key performance indicator (KPI)-based objectives. The KPIs are extracted from S-parameter responses. An all-resonator diplexer with 22 variables and a Tee-junction diplexer with 23 variables are designed using the proposed method via one-off optimization, showing 100% and 80% success rates, respectively.

Electric boilers (EBs) provide an alternative method to deal with the accommodation of curtailed wind power. To pursue the minimum coal consumption in the system, a dispatching model integrating combined-heat-and-power (CHP) plants and EBs in different locations is developed, and the penalty of wind power curtailment and cost of EB employment are also incorporated in the model. The transmission loss and transportation lag of heat-supply network as well as the elasticity of heat load are considered in this paper. A kind of constrained programming with stochastic and fuzzy parameters is applied to deal with the uncertainties. A case in East Inner Mongolia in China demonstrates that the EBs are able to absorb curtailed wind power and supply the heat. The results indicate that the utility of EBs in the primary or secondary heat-supply network to accommodate curtailed wind power is mainly related to the efficiency of heat transmission and the elasticity of heat load.

The noncontact and wide-band characteristics of the ultrasonic signal excited by the laser make it widely used in nondestructive detection. In this paper, the finite element method is used to simulate multiple ultrasonic signal modes and the responses of them to metal defects. The simulation shows that the changes of excited Rayleigh-wave (RW) and shear-wave (SW) can be used to research the surface crack and internal defect of the sample respectively. By analyzing the reflected wave of RW and SW at the inspection point, the effects of surface crack and internal defect on the ultrasonic signals of the two modes are quantitatively compared. The results show that the time delay of the Rayleigh echo has a linear relationship to the width and transverse position of the crack, and the peak-valley difference of Rayleigh echo at the inspection point and the crack depth can also be fitted linearly. In addition, the Y component displacement of the bottom reflected SW also decays with the change of the internal defect. This simulation model provides a theoretical basis for further experimental verification basis.

At present, the primary technical deterrent to the use of strawberry harvesting robots is the low harvest rate, and there is a need to improve the accuracy and real-time performance of the localization algorithms to detect the picking point on the strawberry stem. The pose estimation of the fruit target (the direction of the fruit axis) can improve the accuracy of the localization algorithm. This study proposes a novel harvesting robot for the ridge-planted strawberries as well as a fruit pose estimator called rotated YOLO (R-YOLO), which significantly improves the localization precision of the picking points. First, the lightweight network Mobilenet-V1 was used to replace the convolution neural network as the backbone network for feature extraction. The simplified network structure substantially increased the operating speed. Second, the rotation angle parameter $\alpha $ was used to label the training set and set the anchors; the rotation of the bounding boxes of the target fruits was predicted using logistic regression with the rotated anchors. The test results of a set of 100 strawberry images showed that the proposed model's average recognition rate to be 94.43% and the recall rate to be 93.46%. Eighteen frames per second (FPS) were processed on the embedded controller of the robot, demonstrating good real-time performance. Compared with several other target detection methods used for the fruit harvesting robots, the proposed model exhibited better performance in terms of real-time detection and localization accuracy of the picking points. Field test results showed that the harvesting success rate reached 84.35% in modified situations. The results of this study provide technical support for improving the target detection of the embedded controller of harvesting robots.

Gravitational search algorithm (GSA) inspired from physics emulates gravitational forces to guide particles' search. It has been successfully applied to diverse optimization problems. However, its search performance is limited by its inherent mechanism where gravitational constant plays an important role in gravitational forces among particles. To improve it, this paper uses chaotic neural oscillators to adjust its gravitational constant, named GSA-CNO. Chaotic neural oscillators can generate various chaotic states according to their parameter settings. Thus, we select four kinds of chaotic neural oscillators to form distinctive chaotic characteristics. Experimental results show that chaotic neural oscillators effectively tune the gravitational constant such that GSA-CNO has good performance and stability against four GSA variants on functions. Three real-world optimization problems demonstrate the promising practicality of GSA-CNO.

Density peaks clustering (DPC) algorithm is a novel density-based clustering algorithm, which is simple and efficient, is not necessary to specify the number of clusters in advance, and can find any nonspherical class clusters. However, DPC relies heavily on the calculation methods of the cutoff distance threshold and local density and cannot analyze complex manifold data, especially datasets with uneven density distribution and multiple peaks in the same cluster. To solve these problems, we propose an improved density peaks clustering algorithm based on the layered k-nearest neighbors and subcluster merging (LKSM_DPC). First, we redefine the local density calculation method using the layered k-nearest neighbors. To adapt to datasets with different densities, the k-nearest neighbors are divided into multiple layers. Second, for the multiple peaks in the same cluster problem, we design a new mechanism to calculate the similarity of subclusters based on the idea of shared neighbors and Newton's law of gravitation, and a subcluster merging strategy is proposed. To prove the effectiveness of our algorithm, we compare the LKSM_DPC with K-means, DBSCAN, DPC, and DPC derivatives for 24 datasets. A large number of experiments demonstrate that our algorithm can often outperform other algorithms.

Key frame extraction technology is one of the core technologies of content-based video retrieval. For video types with complex content, various scenes, and rich actions, the performance of existing key frame extraction methods is not ideal. Based on the Visual Geometry Group (VGG), this article proposes an image saliency extraction model assisted by deep prior information, and uses a large-scale data set for training on the server to obtain a trained model, and then integrates multiple features. The saliency extraction algorithm is combined with the image saliency extraction model assisted by deep prior information, and a saliency extraction algorithm based on multi-feature fusion and deep prior information is proposed. A new method for extracting key frames of motion video is introduced in detail. Taking into account that sports videos in real applications are susceptible to interference from various factors, resulting in poor picture quality, this article constructs a new visual attention model for moving targets in sports videos, which integrates images. The combination of multiple features of the bottom-level features and the skin color confidence map of the moving target overcomes the problem that a single feature cannot fully express the moving target. Since the processing object in this article is for the moving target in the video of the sports room, the extracted moving target can provide samples for video post-processing. The experimental results show that the proposed key frame extraction algorithm can quickly grasp the pedestrian information in the motion video and provide effective processing samples for the motion target for video post-processing.

The Solid (Social Linked Data) project focuses on data sharing and privacy security and aims to build a decentralized ecosystem that radically changes the way web applications work today. Our goal is to introduce a “trust access authentication system” to achieve secure authentication and fine-grained access control, thereby promoting the implementation of Solid. Blockchain, equipped with multiple security properties and authentication functions, is a crucial technology. In this paper, we present a blockchain-assisted system for secure authentication in Solid and for implementation of fine-grained access control policies. Specifically, we explore to integrate threshold RSA signatures in a permissioned blockchain system to enable a fault-tolerant distributed signature scheme, thereby enhancing the resilience and robustness of authentication system. Moreover, we utilize smart contract to control transaction flows and manage access control policies automatically. Experimental results show that our proposed trust access authentication system enhances security, scales well, and is efficient and economically feasible.

There is a very important index in the influence of human visual perception, that is, the degree of the motion blur about the moving image on display device. In this paper, a new kind of improved method is proposed to calculate the response time of liquid crystal under the new backlight driving and modulation, which is more accurate than the traditional moving window integration method. This paper also discusses the eye tracking movement of the human visual system, including studying the influence of image content, size, motion speed and motion direction, based on the human vision property study and the analyzes of the simulation model of image motion artifacts perceived by human eyes. Experiments are performed, using the human eye vision perception measurement system. The experiment data shows that the edge blur time can be reduced to 10.5 ms, compared with the 16.5 ms in traditional drive, reduced by 36.4%, and the human perception of motion blur can be markedly reduced. Also, the results show that the simulation results are in good agreement with the actual measurement data by modifying the proportion coefficient of the eye tracking object model to 0.9. At the same time, through the improved human eye tracking model of smooth moving objects, the actual motion blur phenomenon perceived by the human eye can be calculated more accurately, and the results can be used in image processing algorithms to improve the quality of the moving image on the display.

This paper addresses the trajectory tracking problem for constrained high dynamic electro-hydraulic actuator in the presence of time-varying parameters, high frequency external load interference, measurement noise and some unmeasurable states. An adaptive robust optimal control scheme is proposed for the electro-hydraulic actuator in legged robot. The framework of our presented scheme is based on a linear time-varying model predictive controller (LTV-MPC) embedded with a multi-scale online estimator (MEKF). With fast- varying and slow- varying time scales, the MEKF part is used not only for measurable states filtering and unmeasurable states estimation, but also for time-varying parameters and external load interference estimation, which will be integrated into the mpc model in real time. The LTV-MPC part is a trajectory tracking controller designed by constrained MPC with an approximate high-precision real-time model and a rapidly solved cost function, which guarantees that the input and output constraints are satisfied during the receding horizon and optimal control process. Finally, with a series of highly dynamic conditions, the comparison experiment results show that the proposed controller has a simple design process, strong adaptive robust performance and trajectory tracking performance, which verifies the effectiveness of the control scheme.

Online games are the most popular form of entertainment among youngsters as well as elders. Recognized as e-Sports, they may become an official part of the Olympic Games by 2020. However, a long waiting time for matchmaking will largely affect players' experiences. We examine different matchmaking mechanisms for 2v2 games. By casting the mechanisms into a queueing theoretic framework, we decompose the rule design process into a sequence of decision making problems, and derive the optimal mechanism with minimum expected waiting time. We further the result by exploring additional static as well as dynamic rule designs' impacts. In the static setting, we consider the game allows players to choose sides before the battle. In the dynamic setting, we consider the game offers multiple zones for players of different skill levels. In both settings, we examine the value of choice-free players. Closed form expressions for the expected waiting time in different settings illuminate the guidelines for online game rule designs.

A dc solid-state power controller (SSPC) is a semiconductor-based circuit breaker used for fault isolation in aerospace applications. Device selection is important to ensure the reliability and performance of the SSPC. In this paper, Si, SiC, and GaN field-effect transistors are compared under steady-state and transient conditions. Experimental results are presented to evaluate the performance and characteristics. Nanyang Technological University National Research Foundation (NRF) Accepted version This work is jointly funded by the National Research Foundation (NRF) of Singapore, Rolls-Royce Singapore Pte. Ltd, and Nanyang Technological University, Singapore.

Brain storm optimization (BSO) is a swarm intelligence optimization algorithm which is proven to have practical values in various fields. During these years, many modifications have been facilitated to effectively improve BSO's search performance. So far, these modifications focus on improving the solution quality by applying different clustering methods and learning strategies, in which the population diversity is often neglected. However, in recent studies, population diversity plays a more significant role in designing optimization algorithm. A population that maintains its diversity in a high level can easily obtain better solutions than the one with low level of diversity. Therefore, this paper proposes a control method that evaluates the population diversity of BSO to improve its performance. Two diversity measures, which are known as distance-based diversity and fitness-based diversity, are implemented to realize the adaptation of algorithm parameters. The new algorithm is called multiple diversity-driven BSO (MDBSO). Its performance is verified by CEC2017 benchmark function suit and a neuron model training task. The results demonstrate the effectiveness and efficiency of MDBSO.

Micro-expression recognition has been an active research area in recent years, it plays an important role in psychology and public security. Due to the aspects of short duration and subtle movement, it is challenging to extract spatiotemporal features of micro-expressions. The existing methods only extract features in the three-dimensional orthogonal plane and fail to make full use of that information. To solve this problem, we propose a new Local Cubes Binary Patterns (LCBP) method for micro-expression recognition. LCBP is cascaded by the motion information LCBPdirection, the amplitude information LCBPamplitudes, and the spatial information LCBP3D to obtain the spatiotemporal features. The advantage of LCBP is its ability to preserve the spatiotemporal information and the low feature dimension. Furthermore, to increase the discrimination of features in micro-expression sequences, we apply a differential calculation energy map to find regions of interest (ROI) for getting a weighted energy map. The final micro-expression feature acquired by fusing the LCBP features and the weighted energy map are classified through the Support Vector Machine (SVM). We evaluate the proposed method on four published micro-expression databases including SMIC, CASME, CASME2, SAMM. Experimental results demonstrate that our proposed method achieves promising performance for micro-expression recognition.

Seru production can be used to enhance productivity, such as makespan reduction and manpower saving. Seru system operation includes two NP-hard problems, i.e., seru formation and seru scheduling. The exact solution cannot be obtained by solving only one of seru formation and seru scheduling. We develop a cooperative coevolution algorithm for the Seru production with minimizing makespan by solving the seru formation and seru scheduling simultaneously. The cooperative coevolution algorithm includes two evolution algorithms, i.e., the algorithm combining generic algorithm and local search, and the ant colony optimization algorithm. The former algorithm is used to deal with the evolution of seru formation. The latter is used to find a better seru scheduling. In the cooperative mechanism, the two algorithms cooperate to seek a better solution of seru system operation. Finally, extensive-tested experiments show that the proposed cooperative coevolution algorithm can obtain a better solution than all the existing algorithms and, even, can obtain the exact solution for some medium-and-small instances.

Feature fusion is widely used in person re-identification (re-ID) and has been proven effective. However, it is difficult to know which features are effective to identify a specific person and how to fuse features to explore complementary information and apply the advantages of each feature. Motivated by these problems, this paper proposes a new method of person re-ID to fuse the recognition results of multiple features at the rank level. Three innovations are included in this method: first, multiple metric spaces are constructed based on the correlation of different features to generate multiple rank results; second, the most similar candidates in each corresponding rank list is converted into a graph structure by our proposed weighted center graph (WCG), and we use an adaptive value $K$ to automatically seek the most similar images of each query, thus improving the accuracy of candidate targets. Finally, to evaluate the effect of each WCG, a discriminative power coefficient is designed and used to assign a proper coefficient for each WCG according to the discriminative power of corresponding features. The result can be obtained by re-ranking the fused WCG. The extensive experiments on five datasets demonstrate the matching rate of our proposed method by comparing with several state-of-the-art methods. Our code is available at https://github.com/gengshuze/WCG.git .

As a new paradigm of information technology, the Internet of Things (IoT) is attracting increasing attention from various industrial fields. It is foreseeable that the applications of IoT will be prevalent in the public transportation system and bring changes to the system in the near future. In this paper, we analyze the impact of IoT environment on the public transportation system, propose a new framework of the intelligent public transportation system based on IoT, and present the deployment of the elements, the communication network, and the three-tier architecture of the system in detail. We also present the information flow, technical scheme, optimization model, and algorithm of the main modules of dynamic optimization of the system. The innovative points of this paper lie in: (1) a new framework for public transport system based on IoT, which integrates the scheduling problems of subway, bus, and shared taxi, is proposed for better-coordinated transfer solutions; (2) transport flow prediction methods based on periodic patterns mining is proposed for road flow analysis and passenger flow analysis, and; (3) mathematical model and DSS-based evolutionary computation algorithm are proposed for solving the dynamic bus scheduling and controlling problems. The proposed intelligent transport system based on IoT can assist the decision makers to increase the utilization rate of the transport resources, improve the efficiency of scheduling, and reduce passengers’ traveling time.

In this paper, a third-order waveguide bandpass filter (BPF) based on slotted spherical resonators with a wide spurious-free stopband is presented. The resonator consists of a spherical cavity with slots opened at the top and bottom. Compared with a non-slotted spherical resonator, the slotted resonator suppresses the two spurious modes (TM211 and TE101) whilst maintaining the fundamental TM101 mode. The unloaded quality factor of the TM101 mode is not significantly degraded. This is achieved by interrupting surface current and radiating the unwanted spurious modes with the slots. The BPF is designed at a center frequency of 10 GHz with a fractional bandwidth of 1%. Two filter prototypes are fabricated, one using metal-based selective laser melting (SLM) and the other by polymer-based stereolithography apparatus (SLA) techniques. The slots also facilitate the copper electroplating process for the SLA-printed filter. The measured results show that the average passband insertion losses of the SLM- and SLA-printed filters are 0.33 and 0.2 dB, respectively. The corresponding passband return losses are better than 22 and 20 dB. The filters demonstrate excellent passband performance and wide spurious-free stopbands up to 16 GHz with stopband rejections of over 20 dB.

In this paper, we theoretically and experimentally investigate the all-optical modulation characteristics of a microfiber coupler combined Sagnac loop (MCSL). We inject pump lights with different intensities into the MCSL and utilize the thermo-optical effect from the microfiber coupler to control the phase relationship between the counter-propagating beams inside the Sagnac loop. The spectral characteristics from the Sagnac loop can be controlled accordingly, and the rate of the frequency comb (interference characteristic wavelength) shift can be 1 pm/mW. In addition, at a specific wavelength (e.g., 1550 nm), the MCSL can be used as an intensity modulator as well as a reflectivity-tunable Sagnac reflector. The reflectivity variation of the Sagnac reflector can reach about 10% when the pump light is at milliwatt level. The intensity modulation depth can be 9.3% at 100 Hz when the intensity-modulated 980 nm pump light imposed on the MCSL is 5.5 mW. Based on the thermo-optical effect, the MCSL can achieve the all-optical intensity modulation and all optically tunable filtering functions. The device is expected to play an important role in the development of tunable, multi-wavelength, and compact fiber lasers.

With the development of the mobile Internet, novel things are constantly emerging, and among them, online car-hailing is one of the representatives. However, the rapid development of online car-hailing also brings about normative problems and safety loophole, the most prominent of which is the inconsistency between operating cars and registered drivers. Some existing algorithms based on curve similarity measurements, such as Hausdorff distance and discrete Fréchet distance, can be used to solve this problem. However, they only consider one of the characteristics of the two curves, such as the distance or the area between two curves, which does not achieve good results on this problem. On this ground, we propose a model based on curve comparison, named multi-feature fusion (MFF), which extracts the features of length, distance, and area from the GPS positioning track of car and the application mobile phone of the driver, to testify whether the car is being operated by the registered driver and thus solving the problem of mismatch between the driver and the car during the operation. Among them, the MFF model designs different algorithms for different features and fuses different features by an ensemble learning method. The experimental results prove that the model can effectively detect the mismatch between drivers and cars.

© 2019 IEEE. The key to dynamic pressure calibration is the generation of sinusoidal pressure signals. This paper uses a liquid-gas two-phase piston sinusoidal pressure generator for generating higher frequency pressure signals. The design of the contraction section has a large effect on the sinusoidal pressure generator. The flow field characteristics of the contraction section were simulated by Fluent simulation software. We analyze the frequency response, exit velocity distribution and axial pressure distribution of different contraction curves, then select the appropriate contraction curve and verify it by experiment. The simulation work in this paper provides a reference for the design of liquid-gas two-phase high-frequency sinusoidal pressure generator.

Multiple LiDARs have progressively emerged on autonomous vehicles for rendering a rich view and dense measurements. However, the lack of precise calibration negatively affects their potential applications. In this paper, we propose a novel system that enables automatic multi-LiDAR calibration method without any calibration target, prior environment information, and manual initialization. Our approach starts with a hand-eye calibration by aligning the motion of each sensor. The initial results are then refined by an appearance-based method by minimizing a cost function constructed by point-plane distance. Experimental results on simulated and real-world data demonstrate the reliability and accuracy of our calibration approach. The proposed approach can calibrate a multi-LiDAR system with the rotation and translation errors less than 0. 04rad and 0. 1m respectively for a mobile platform.

© 2019 IEEE. Falls happen frequently to older people in their daily life. Heel and toe clearances are regarded as significant indicators for the risk of falling. In this study, we developed a foot-worn inertial system to capture kinematic gait data. An estimation method of continuous heel and toe clearances during walking was proposed. Based on heel strike and toe off events detection, a foot movement model was presented to calculate the continuous heel and toe clearances. To compensate the sensor drift, an algorithm combined Zero Velocity Update, Zero Angular Rate Update, Zero Vertical Displacement Update plus a liner drift correction model was implemented. To verify the effectiveness of the drift compensation algorithm, the calculated IMU clearances before and after the implementation of the algorithm were compared and the walking tests results supported its feasibility. Furthermore, the continuous heel and toe clearances were estimated successfully using the foot movement model.

© 2019 IEEE. The three-axis MEMS magnetometer is widely applied to assist the inertial sensor to estimate the attitude (especially heading) in strapdown inertial navigation system. It is necessary that the system errors of the magnetometer are analyzed and compensated, just like the inertial sensor. The system errors that contain biases, scale factors, nonorthogonal angles and iron effects can be quantified by the calibration. In general, error items can be derived based on the geomagnetic field vector observation. In most magnetometer calibration algorithm, the data from the sensor is processed directly by ellipsoid fitting without any pretreatment. In order to guarantee the reliability of the data, this paper proposes a new method for testing the rationality and gross errors of the calibration data based on geometric parameters of the ellipsoid. The experimental results show that the unreasonable data is detected, so that the calibration accuracy is improved to some extent.

Multiple lidars are prevalently used on mobile vehicles for rendering a broad view to enhance the performance of localization and perception systems. However, precise calibration of multiple lidars is challenging since the feature correspondences in scan points cannot always provide enough constraints. To address this problem, the existing methods require fixed calibration targets in scenes or rely exclusively on additional sensors. In this paper, we present a novel method that enables automatic lidar calibration without these restrictions. Three linearly independent planar surfaces appearing in surroundings is utilized to find correspondences. Two components are developed to ensure the extrinsic parameters to be found: a closed-form solver for initialization and an optimizer for refinement by minimizing a nonlinear cost function. Simulation and experimental results demonstrate the high accuracy of our calibration approach with the rotation and translation errors smaller than 0.05rad and 0.1m respectively., Comment: 6 pages, 8 figures, accepted by 2019 IEEE Intelligent Vehicles Symposium (IVS)

In many THz Time Domain Spectroscopy configurations, the separate acquisition of background and sample measurements have significant impact on data quality due to the need to perform subsequent measurements as well as the change in beam coupling geometry due to the removal of the sample in the reference measurement. In this short paper, we report on the design and build of a dual-polarization TDS system to perform synchronous background/sample measurements. This is achieved through the splitting of the PCA-generated THz beam with a THz polarizer to obtain two orthogonal polarized beams which are balanced and modulate the indices of two separate crystals on which the probe beam (also split) is reflected off before being detected. The two branches of the system show good similarity, and can be used to perform background and sample measurements synchronously, thus not being affected by changes in temperature and humidity, as well as small fluctuations in source power. Small differences in detector optical efficiency can be calibrated out at the start of a measurement session.

In this paper, we address the energy storage management problem in distribution networks from the perspective of an independent energy storage manager (IESM) who aims to realize optimal energy storage sharing with multi-objective optimization, i.e., optimizing the system peak loads and the electricity purchase costs of the distribution company (DisCo) and its customers. To achieve the goal of the IESM, an energy storage sharing strategy is therefore proposed, which allows DisCo and customers to control the assigned energy storage. The strategy is updated day by day according to the system information change. The problem is formulated as a bi-level mathematical model where the upper level model (ULM) seeks for optimal division of energy storage among Disco and customers, and the lower level models (LLMs) represent the minimizations of the electricity purchase costs of DisCo and customers. Further, in order to enhance the computation efficiency, we transform the bi-level model into a single-level mathematical program with equilibrium constraints (MPEC) model and linearize it. Finally, we validate the effectiveness of the strategy and complement our analysis through case studies.

One of the essential problems on Bayesian networks (BNs) is parameter learning. When purely data-driven methods fail to work, incorporating supplemental information, like expert judgments, can improve the learning of BN parameters. In practice, expert judgments are provided and transformed into qualitative parameter constraints. Moreover, prior distributions of BN parameters are also useful information. In this paper we propose a Bayesian approach to learn parameters from small datasets by integrating both parameter constraints and prior distributions. First, the feasible parameter region is derived from constraints. Then, using the prior distribution, a posterior distribution over the feasible region is developed based on the Bayes theorem. Finally, the parameter estimations are taken as the mean values of the posterior distribution. Learning experiments on standard BNs reveal that the proposed method outperforms most of the existing methods.