会议名称:
IEEE International Conference on Fuzzy Systems (FUZZ-IEEE) / IEEE World Congress on Computational Intelligence (IEEE WCCI) / International Joint Conference on Neural Networks (IJCNN) / IEEE Congress on Evolutionary Computation (IEEE CEC)
会议时间:
JUL 18-23, 2022
会议地点:
Padua, ITALY
会议主办单位:
[Bai, Jun;Sajjanhar, Atul;Xiang, Yong] Deakin Univ, Sch Informat Technol, Geelong, Vic, Australia.^[Tong, Xiaojun] Wuhan Text Univ, Sch Comp Sci & Artificial Intelligence, Wuhan, Peoples R China.^[Zeng, Shan] Wuhan Polytech Univ, Sch Math & Comp Sci, Wuhan, Peoples R China.
会议论文集名称:
IEEE International Joint Conference on Neural Networks (IJCNN)
关键词:
distributed learning;federated learning;data heterogeneity;Non-IID data;heterogeneous model fusion
摘要:
Federated Learning (FL) offers a novel distributed machine learning context whereby a global model is collaboratively learned through edge devices without violating data privacy. However, intrinsic data heterogeneity in the federated network can induce model heterogeneity, thus posing a great challenge to the server-side model aggregation performance. Existing FL algorithms widely adopt model-wise weighted averaging for client models to generate the new global model, which emphasizes the importance of the holistic model but ignores the importance of distinctions between internal parameters of various client models. In this paper, we propose a novel parameter-wise elastic weighted averaging aggregation approach to realize the rapid fusion of heterogeneous client models. Specifically, each client evaluates the importance of model internal parameters in the model update and obtains the corresponding parameter importance coefficient vector; the server implements the parameter-wise weighted averaging for each parameter based on their importance coefficient vectors, thereby aggregating a new global model. Extensive experiments on MNIST and CIFAR-10 datasets with diverse network architectures and hyper-parameter combinations show that our proposed algorithm outperforms the existing state-of-the-art FL algorithms on the performance of heterogeneous model fusion.
摘要:
In recent years, various deep learning based methods have been successfully developed for change detection, such as Convolutional Neural Network (CNN) based U-Net and its variants, and Transformer based ones. However, CNNs lack the ability to effectively learn global representations, while Transformers neglect to learn local representations. Therefore, in this paper we propose a novel deep network, namely Multi-scale Attention based Transformer U-Net (MATU), to take advantages of CNNs and Transformers for learning both local and global features effectively. The backbone of our proposed MATU is a U-Net. In the encoder, a Siamese network is used to extract features from two input images, which is followed by a transformer module to further refine the feature pairs produced by the Siamese network. The difference of the refined feature pairs is fed into an Atrous Spatial Pyramid Pooling (ASSP) module to generate a distance map. Moreover, axial-attention blocks are integrated in the decoder with the corresponding multi-level feature differences of the encoder to progressively produce and improve the change map through attention upsampling. Extensive experiments on two widely used benchmark datasets SYSU-CD and LEVIR-CD demonstrate that the proposed MATU method achieves the state-of-the-art performance. Our code is available at https://github.com/easm002/MATU.
关键词:
Hyperspectral unmixing;weight-sharing architecture;stick-breaking process;Dirichlet distribution
摘要:
Recently, the learning-based method has received much attention in the unsupervised hyperspectral unmixing, yet their ability to extract physically meaningful endmembers remains limited and the performance has not been satisfactory. In this paper, we propose a novel two-stream Dirichlet-net, termed as uTDN, to address the above problems. The weight-sharing architecture makes it possible to transfer the intrinsic properties of the endmembers during the process of unmixing, which can help to correct the network converging towards a more accurate and interpretable unmixing solution. Besides, the stick-breaking process is adopted to encourage the latent representation to follow a Dirichlet distribution, where the physical property of the estimated abundance can be naturally incorporated. Extensive experiments on both synthetic and real hyperspectral data demonstrate that the proposed uTDN can outperform the other state-of-the-art approaches.
摘要:
This paper presents a two-stage multi-objective evolutionary algorithm based on classified population (TSCEA) to solve vehicle routing problem with time windows (VRPTW). It is a well-known NP-hard discrete optimization problem with three objectives: to minimize the total distance cost, to minimize the number of vehicles, and to optimize the balance of routes within a limited time. For TSCEA, there are two stages: In the first stage, a population is explored using the proposed algorithm and then classified according to the number of vehicles, we call this process population classification; In the second stage, Pareto solution set of tri-objective VRPTW is obtained by optimizing the classified population again. The advantages of classified population structure are that for the first stage, this population that the number of vehicles of each individual is in this range composed of the upper and lower bounds of vehicles can be classified as different small populations with the same number of vehicles. Due to the evolution of small population, Pareto solution set with better extensibility can be searched. For the second one, it can reduce the dimension of tri-objective function, that is, three objective functions can be reduced to two objective functions because one of them has been identified in the first stage. Moreover, to resolve the nonlinear discrete problems, the computational approach of crowding degree is modified. The paper chooses Solomon benchmark instances as testing sets and the simulated results show that TSCEA outperforms the compared algorithms in terms of quality or extension, which verified the feasibility of the algorithm in solving tri-objective VRPTW.
摘要:
Traffic network transportation optimization problem (TNTOP) has important applications in logistics distribution fields. In various disciplines, methods about the solutions-termed TNTOP can have shown promising performance from different types of detection, at different conditions. Due to the limitatioins of the calculation speed of traditonal algorithms, it is rare that a simple unmodified method provides complete techniques of tackling large-scale TNTOP. We use the term P systems to solve the above limitatioins. Specifically, it is a tissue-like P system with four cells based on particle swarm algorithm, referred to as MPSO. In this system, the modified prim algorithm and the position-updated mechanism are adopted to generate and update all particle individuals, velocity-updated mechanism and an exchange-tree strategy are adopted to balance exploration and exploitation processes. Besides, some special strategies are also added to this systems. Numerous experiments are presented to verify the performance of the MPSO. The results show that it can generate the individuals of higher quality in shorter computation time when comparing to other benchmark algorithms. These empirical results validate the effectiveness and competitiveness of our proposed algorithm in solving TNTOP in terms of both quality and speed.
作者机构:
[Zhao, Juanjuan; Liu, Changhua] Wuhan Polytech Univ, Dept Math & Comp, POB 430023, Wuhan, Hubei, Peoples R China.
会议名称:
5th Annual International Conference on Information System and Artificial Intelligence (ISAI)
会议时间:
MAY 22-23, 2020
会议地点:
Zhejiang, PEOPLES R CHINA
会议主办单位:
[Zhao, Juanjuan;Liu, Changhua] Wuhan Polytech Univ, Dept Math & Comp, POB 430023, Wuhan, Hubei, Peoples R China.
会议论文集名称:
Journal of Physics Conference Series
摘要:
According to the "Top Ten Security Vulnerabilities List" (OWASPTop 10) released by OWASP in 2017, SQL injection attacks are still at the top of the list, and there are many ways of SQL injection attacks, which cause great harm. Although there are many vulnerability scanning tools, there is still a high rate of false negatives. Aiming at the current problems of SQL injection vulnerability detection, this paper proposes a scanning tool for SQL injection vulnerabilities. First, use the crawler framework scrapy to obtain the URL associated with the form and the a tag, and segment the URL based on the improved simhash algorithm. Deduplicate the link, then analyze the injection point to modify the URL parameter value injection test, and determine whether there is a vulnerability based on the response result of the server. The experimental results show that the detection method achieves a 96.50% URL deduplication rate in the crawler module, which greatly reduces the rate of false negatives. It is more suitable for detecting whether a website has a SQL injection vulnerability.
会议名称:
11th International Symposium on Multispectral Image Processing and Pattern Recognition (MIPPR) - Parallel Processing of Images and Optimization Techniques; and Medical Imaging
会议时间:
NOV 02-03, 2019
会议地点:
Wuhan, PEOPLES R CHINA
会议主办单位:
[Yang, Minjun;Hong, Juan;Tang, Xiaoyue;Lee, Yaqin] Wuhan Polytech Univ, Sch Math & Comp Sci, Wuhan, Hubei, Peoples R China.
会议论文集名称:
Proceedings of SPIE
关键词:
Median Filter;Morphology;Gaussian Smoothing;Refinement algorithm;Detect Lumbar Disc Degeneration
摘要:
Recently, diagnosis, therapy and monitoring of human diseases involve a variety of imaging modalities, such as magnetic resonance imaging (MRI), computed tomography (CT), Ultrasound (US) and Positron-emission tomography (PET) as well as a variety of modern optical techniques. The degeneration of lumbar intervertebral disc has become a common disease in modern society. Currently, the most commonly used method is the diagnostic grade based on MRI technology, among which Pfinmann grading system is most widely used in clinic. The Pfinmann grading system is mainly based on the measurement of the average height of the lumbar intervertebral disc and the intensity of the signal of the nucleus pulposus and the inner and outer edge of the fiber ring in MR images. With the degeneration of the intervertebral disc, the signal of the inner and outer edge of the annulus also decreases, so the error caused by the method of measuring the average height of the lumbar intervertebral disc is larger. Therefore, we proposed an algorithm based on morphology to detect lumbar intervertebral disc in MRI spinal images. First, the median filter is used to remove noise in MRI and then the lumbar intervertebral disc is extracted through morphological processing. Then, the image is smoothed by combining with gaussian filtering. Finally, the result map of lumbar intervertebral disc is obtained and its area is calculated. In the analysis and comparison of the detection results of the lumbar intervertebral disc, the skeleton extraction diagram of the detection results of the lumbar intervertebral disc was obtained after processing the image of the detection results of the lumbar intervertebral disc with the thinning algorithm. According to the analysis, the degree of laminar disc skeleton and upper and lower vertebral body is as high as 90%. This paper also briefly introduces the application direction of this measurement algorithm in medicine: 1. Improve doctors' ability to detect early lumbar disc degeneration. 2. Assist doctors to observe postoperative recovery of patients.
会议名称:
11th International Symposium on Multispectral Image Processing and Pattern Recognition (MIPPR) - Automatic Target Recognition and Navigation
会议时间:
NOV 02-03, 2019
会议地点:
Wuhan, PEOPLES R CHINA
会议主办单位:
[Xu, Xiangrui;Li, Yaqin;Gao, Yunlong;Yuan, Cao] Wuhan Polytech Univ, Sch Math & Comp Sci, Wuhan, Hubei, Peoples R China.
会议论文集名称:
Proceedings of SPIE
关键词:
Deep neural network;identity number (ID);Ownership verification
摘要:
Amid the maturity of machine learning, deep neural networks are gradually applied in the business sector rather than be restricted in the laboratory. However, its intellectual property protection encounters a significant challenge. In this paper, we aim at embedding a unique identity number (ID) to the deep neural network for model ownership verification. To this end, a scheme of generating DNN ID is proposed, which is the criterion for model ownership verification. After embedding, the model can complete the original performance and own a unique ID of this model as well. DNN ID can only be generated by the owner to check the model authorship. We evaluate this method on MNIST. Experiment results demonstrate that the DNN ID can accurately verify the ownership of our trained model.
摘要:
When the model begins a new task, the challenge of naming the "catastrophic forgetting" limits the scalability of the deep learning network, which quickly forgets the learning capabilities it has. The fine-tuning method recommends that the original feature extraction be retained to extract the features of the new task and to achieve the purpose of learning the new class. However, this method degrades performance on previously learned tasks because the shared parameters change without new guidance for the original task-specific prediction parameters. This paper proposes general fine-tune method to reduce catastrophic forgetting in sequential task learning scenarios. The critical idea of the method is fine-tuning the parameters in each layer, unlike the traditional fine tuning only for the last layer. The experimental results show that the new method is superior to fine-tune, in the accuracy of the old task and the performance of the new task is better than that of the EWC. A distinct advantage is that old tasks do not limit the performance of new tasks but provide some support for new tasks.
摘要:
Gaussian mixture model (GMM) can estimate not only the abundances and distribution parameters but also distinct end-member set for each pixel. However, the traditional GMM unmixing model only has proper smoothness and sparsity prior constraints on the abundances and thus cannot excavate the local spatial information in hyperspectral image (HSI). Thus, we propose a new unmixing method with superpixel segmentation (SS) and low-rank representation (LRR) based on GMM called GMM-SS-LRR, which can consider the local spatial correlation of HSI. First, we adopt the principal component analysis (PCA) to obtain the first principal component of HSI, which contains the most information for the entire HSI. Then, we adopt the SS in the first principal component of HSI to obtain the homogeneous regions, and the abundances in each homogeneous region have the underlying low-rank property. Finally, we unmix the pixels in each homogeneous region of HSI depending on the low-rank property of abundances. Experiments on synthetic datasets and real HSIs demonstrate that the proposed GMM-SS-LRR is efficient compared with other current popular methods.
关键词:
The polarizing beam splitter;multimode interference coupler;Mach-Zehnder interference
摘要:
The polarizing beam splitter (PBS) based on asymmetric Mach-Zehnder (MZ) interferometer is described theoretically by means of transfer matrix. And the intensity formulas of the output port are obtained according to the theoretical description. The width and length of the PBS interference arm are determined by ways of theoretical calculation and simulation design, and it is found that the theoretical calculation is in good agreement with the simulation design results. The final simulation results show that the designed PBS has a good performance, and the extinction ratio is higher than 30dB in the whole C band. It is expected to realize polarization splitting function in optical fiber communication system. The asymmetric MZ PBS is expected to be widely used in optical fiber communication systems that require polarization splitting.
作者机构:
[Gao, Zunhai] Wuhan Polytech Univ, Sch Math & Comp Sci, Wuhan, Hubei, Peoples R China.;[Chen, Zhuo] Wuhan Polytech Univ, Sch Econ & Management, Wuhan, Hubei, Peoples R China.
会议名称:
2019 2nd International Conference on Mechanical Engineering, Industrial Materials and Industrial Electronics (MEIMIE 2019)2019年第二届机械工程、工业材料和工业电子国际会议(Meimie 2019)
会议时间:
2019-03-29
会议地点:
大连
会议主办单位:
[Gao, Zunhai] Wuhan Polytech Univ, Sch Math & Comp Sci, Wuhan, Hubei, Peoples R China.^[Chen, Zhuo] Wuhan Polytech Univ, Sch Econ & Management, Wuhan, Hubei, Peoples R China.
会议论文集名称:
2019 2nd International Conference on Mechanical Engineering, Industrial Materials and Industrial Electronics (MEIMIE 2019)2019年第二届机械工程、工业材料和工业电子国际会议(Meimie 2019)论文集
关键词:
graph;edge-chain matrix;multiplication of edge-chain matrix;Eulerianian path;Eulerianian cycle)
摘要:
The initial edge-chain matrix and general edge-chain matrix of graph are presented. The operations of the general edge-chain matrices are derived, by which a method to find all Eulerian cycles is obtained. Only through some power operations of the initial edge-chain matrix, can reveal all Eulerianian cycles which are showed in the final edge-chain matrix. This method can determine whether Eulerianian cycles exist or not and if they do can also find out all of them. It is effective to directed or undirected finite graph. And it can be simplified by computations of some row vectors and column vectors of some power of the initial edge-matrix. This pure mathematical method shows the results more intuitive and makes program operation easier.
摘要:
In video foreground detection, the frame difference method bears a fast detection speed and strong timeliness. However, the detection foreground target is not complete enough, tending to result in voids and poor robustness. Although the Gaussian mixed model does well in detection, a ghost image is easily brought forth when it starts foreground target detection towards sudden motion. As far as these problems are concerned, a foreground detection algorithm based on improved Gaussian mixed model is proposed in this paper. The foreground region detected by the Gaussian mixed model is matched with the one detected by the improved three-frame difference method with the matched foreground reserved. Then the unmatched one is regarded as a 'ghost' region. The background model of the region is updated and the mean of the maximum weighted Gaussian model is replaced in an usage of the pixel of the corresponding area, thus breaking the obstacles in traditional method for detecting holes and 'ghosting' problems. The experimental results have shown that the proposed algorithm has better robustness and accuracy in different backgrounds, and the precision and recall excel that of traditional algorithms.
期刊:
Lecture Notes in Electrical Engineering,2018年474:727-733 ISSN:1876-1100
通讯作者:
Ruan Ling
作者机构:
[Liu Changhua; Wang Yuling; Ruan Ling] Wuhan Polytech Univ, Coll Math & Comp Sci, Wuhan 430023, Hubei, Peoples R China.
通讯机构:
[Ruan Ling] W;Wuhan Polytech Univ, Coll Math & Comp Sci, Wuhan 430023, Hubei, Peoples R China.
会议名称:
12th KIPS International Conference on Ubiquitous Information Technologies and Applications (CUTE) / 9th International Conference on Computer Science and its Applications (CSA)
会议时间:
DEC 18-20, 2017
会议地点:
Taichung, TAIWAN
会议主办单位:
[Ruan Ling;Liu Changhua;Wang Yuling] Wuhan Polytech Univ, Coll Math & Comp Sci, Wuhan 430023, Hubei, Peoples R China.
会议论文集名称:
Lecture Notes in Electrical Engineering
关键词:
Ant colony algorithm;Network failure;Failed path;Pheromone concentration
摘要:
Information-Centric Network (ICN) is being regarded as a promising technology in future Internet. Content Centric Network (CCN) has been proposed to facilitate users acquire multimedia contents. It not only replaces traditional IP-based connections with content delivery but also realizes the features of named data and content store. However, existing cache scheme in CCN does not elaborate on content attributes. The unnecessary content seriously occupies the content store of routers and wastes storage resources. In this paper, we aim to tackle the cache problem of popular and normal contents in CCN. A cache scheme as well as the cache scheme based on content popularity and user locality (CSCPUL) is proposed to solve the cache problem. Results showed that the CSCPUL is superior to traditional cache scheme in CCN in terms of cache cost and cache hit rate.
摘要:
Market economics can achieve optimal allocation of resource by the equilibrium theory, cloud resource provision mechanism based on market economics is studied. For increasing collective revenue further and improving the efficiency and fairness of resource allocation, a resource provision algorithm RPABG is proposed in bargaining market of cloud computing. RPABG builds the model with bargaining game and its objective is to find Nash bargaining solution (NBS). NBS is solved respectively in two different conditions by our proposed iteration algorithm. Simulation experimental results show that RPABG not only has a faster convergence speed, compared with our earlier work based on non-cooperative game, but can improve the overall utility of resource providers and realize Pareto efficiency, which can lead to an optimal allocation of cloud resource with fairness, rationality and equilibrium.
摘要:
Tasks scheduling problem is the key challenge in cloud computing system. For reducing the execution cost of workflow tasks scheduling under the deadline and the budget constraint, a workflow tasks scheduling algorithm based on genetic algorithm in cloud computing is proposed. In our algorithm, each task is assigned priority by an top-down leveling method. By this top-down leveling method, all workflow tasks are divided into the different levels, which can promote the parallel execution of workflow tasks. When code the solution of tasks scheduling, we design a two dimension coding method. And, we design a new genetic crossover and mutation operation to produce new different offsprings for increasing the population diversity. Through the fitness function synchronously considering the scheduling time and the scheduling cost, we can evaluate the individual fitness of population. Through the simulation experiments, we evaluate the performance of our algorithm based on realistic workflows model. The results show that our algorithm has a better performance in reducing the workflow scheduling cost.