摘要:Geometry calibration is a key step to the seamless display of an autostereoscopic tiled display system using multiple projectors,but the robust detection of feature points is more important because the geometry calibration is based on the feature-point detection.Due to the anisotropic refection characteristics of lens sheets for an autostereoscopic display screen,it is very difficult to find a threshold for the threshold-based method to segment those feature images well.In this paper,we propose a robust feature-point detection method for the autostereoscopic tiled display system using multiple projectors.It first detects feature points with the circle detection method based on Hough transform,then,the convex hull of the feature points is computed;undetected feature points are added into the feature-point mesh,and the coordinates of the feature points are updated twice by the linear least square fitting to minimize the errors;and last,based on the feature-point mesh,the geometry calibration is performed for the autostereoscopic tiled display system using multiple projectors.The experimental results show that our feature-point detection approach is robust and precise enough for the geometry calibration.
关键词:feature point detection;Hough transform;circle detection;linear least square fitting;geometry calibration
WANG Xiao-dong, HUO Guang-cheng, SUN Hai-yan, MENG Xiang-xu, SUN Yan-qiang
Vol. 38, Issue 8, Pages: 1736-1740(2010)
摘要:Based on full network coding,opportunistic routing can be applied to solve the problem of space reuse and redundant packet copies in mobile ad hoc networks.Large average delay and bursty arrival flow of packets make the routing defficient.An opportunistic routing based on partial network coding (OR-PNC) is proposed in this paper to encode packets using code vectors with variable length.Experimental results show that OR-PNC can effectively decrease packet average delay by 26% at most and make the arrival packet flow at destination nodes more smoothly.
CHEN Ping, HAN Hao, SHEN Xiao-bing, YIN Xin-chun, MAO Bing, XIE Li
Vol. 38, Issue 8, Pages: 1741-1747(2010)
摘要:In recent years,Integer bugs have been rising sharply and become a potential threat as it is often hidden behind other bugs.In this paper,we propose a tool which can automatically detect Integer bugs.We implement the tool based on static and dynamic program analysis.In the static phase,the tool decompiles a binary and creates the suspect instruction set.In the dynamic phase,it monitors the instructions in the suspect set and generates the test cases to further detect which instructions are real Integer bugs.Our tool has two advantages.First,it provides more accurate and sufficient type information.Second,static analysis reduces the instructions which are monitored at runtime.Experimental results shows that our tool can efficiently detect the Integer bugs in several real-world programs.In addition,our tool has no false negatives and low false positives.
MENG Peng, HUANG Liu-sheng, CHEN Zhi-li, YANG Wei, YANG Ming
Vol. 38, Issue 8, Pages: 1748-1752(2010)
摘要:Translation-Based Steganography (TBS) is a class of text steganography.When translating the same sentence from a natural language to another by different translators,there are normally different translation results.Making use of this property,TBS chooses from different translation results of a sentence according to secrete messages to generate stego-text.Because TBS basically preserves syntactic correctness and semantic coherence,it is comparatively difficulty for traditional detections to detect this method.Our research shows that there exists a detection method for TBS when we know the Machine Translation (MT) set that being used by TBS.So it is crucial important to keep a secret of the MT set.This paper theoretically analyzes the effective of our detection method and gives the detection process and experiment results.Additionally,we propose a method to improve the security of TBS.
关键词:steganography;steganalysis;machine translation;degree of machine reversibility;degree of machine preference;degree of machine generated
摘要:There are some faults in the existing role-based delegation models:(1) most of these models can rarely support a fine delegation granularity effectively;(2) the propagation of permissions in them is poorly-controlled.This paper analyzed these problems,and proposed a fine-grained role delegation model based on mapping mechanism (RDBMPM),which mainly consisted of a vectorizing operator and a measuring operator.Based on these two operators’ concerted computing,the paper introduced the concept of measuring role,which acts as the authorized granularity in the following discussion of delegation constraint mechanism.Finally the significance of the research on RBAC(Role-Based Access Control) delegation model based on mapping mechanism is justified by the simulating of three typical fine-grained delegation models in RDBMPM.
关键词:access control;role based delegation model;mapping mechanism;measuring role
摘要:This paper proposes a new active queue management algorithm named REDu that excavates in depth the essential differences between non-adaptive and adaptive flows. Taking use of information like CHOKe Hit and RED Drop,this algorithm preselects non-adaptive flows and utilizes a heat increasing and decreasing mechanism to compute "heat",a new kind of partial flow state,achieving the detection and punishment of non-adaptive flows. Simulation results based on ns-2 show that,compared with several other active queue management algorithms,REDu can detect and punish non-adaptive flows more precisely,bring more adequate protection to adaptive flows and improve network robustness significantly.
摘要:An energy and location aware ACO based routing algorithm (ELACO) for WSNs is also designed by fusing the energy and location information of nodes.To bypass dead ends,a backward routing mechanism is established,as a result,the routing success rate increases.Simulation results show that the ELACO has a higher search success rate,efficiently balance node energy consumption to prolong the network lifetime.
LIU Xu, LI Jin-bao, JI Shou-ling, GUO Long-jiang, HUI Li
Vol. 38, Issue 8, Pages: 1770-1775(2010)
摘要:A cluster head selecting algorithm based on linear programming called LPCHS and a cluster head sceduling algorithm based on cluster head time ratio called CHSA are proposed to maximize the networks lifecycle.LPCHS constructs a LP equation based on the data flow conservation,energy constraint etc,and obtains the cluster lifecycle,the time of a node being a cluster head and the time ratio correspondingly.CHSA can fulfill the data forwarding among clusters by multi-path routing technology,and obtains a cluster head scheduling scheme based on the cluster head time ratio.
ZHANG Chuang, CHI Jian-nan, ZHANG Zhao-hui, WANG Zhi-liang
Vol. 38, Issue 8, Pages: 1776-1783(2010)
摘要:For the disadvantages of color bilateral noise filtering,this paper presents an integrated approach to address these issues based on edge detection and bilateral filter.First,a new color edge detection algorithm of adaptive partition is derived from cellular neural network (CNN) model.This algorithm inherits the advantage of accurate positioning of CNN gray edge detection,and can deal directly with color images.Next,an edge filter is proposed for image enhancement,which satisfies different stages of denoising through two types of edge detection.On this basis,removing noise of color images uses of the improved bilateral filter.The scope of noise is localized by non-anti-noise edge map to reduce the denoising process and image blur.Average mode of bilateral filter is improved to reduce the weight of noise and the impact of high-frequency noise.Finally,color images are enhanced according to the edge map after filtering.The experimental results show that the proposed approach can effectively reduce the noise while preserving and enhancing edges.
摘要:Accurate and automatic brain tissue segmentation of magnetic resonance (MR) images is a challenging problem because of partial volume (PV) effects,intensity non-uniformity (INU,also known as bias field) and noise.We present an efficient and accurate,fully automatic two-dimensional (2D) and three-dimensional (3D) algorithm for segmenting brain MR images into anatomical major tissue classes such as white matter (WM),gray matter (GM) and cerebrospinal fluid (SCF).Our algorithm is formulated by proposing an objective function based on standard FCM algorithm with bias field correction and neighborhood constrain.In our algorithm,a parameterized model is adopted to express the INU and a neighbor constrain similar to Markov random field (MRF) is proposed to express spatial consistency of brain tissue.Experimental results with both synthetic and real data are included,as well as comparisons of the performance of our algorithm with that of other published methods.The validation of the algorithm shows good accuracy and fast convergence.
关键词:image segmentation;bias field correction;neighborhood constrains;fuzzy-clustering;brain tissues
摘要:We proposed to describe the relationship between the pixel grey values of the face images under frontal and non-frontal illumination conditions by using a second order polynomial model.Correspondingly,an illumination normalization method based on such nonlinear model was formed.The proposed method learns the illumination variations in a statistical manner by using regression model without any prior physical knowledge.Furthermore,in order to improve the visual quality,a PCA-based weighting compensation for the normalized face image was proposed.The experimental results on the Extended Yale B and CMU-PIE face databases show that the proposed method can attain good visualization for face images,and significantly improve the face recognition performance.
关键词:face recognition;illumination normalization;second order polynomial mode
SHU Guang, YAO Li-xiu, YANG Xiao-chao, ZUO Xin, YANG Jie
Vol. 38, Issue 8, Pages: 1798-1802(2010)
摘要:Facial caricaturing from images could be widely used in digital entertainment and emotional communication.Previous methods mainly focused on 2D caricature generation.Although the Morphable Model can synthesize 3D faces with various attributes,it is not fit for the real-time application as the high computational cost.We proposed an improved 3D caricaturing method based on Sparse Morphable Model.It is much faster and only requires a single image.Firstly a sparse face model is reconstructed from the image by Sparse Morphable Model.Then a General Face Model is warped and texture is synthesized.Finally the 3D model is caricatured.Experimental results demonstrate that this method can generate impressive 3D caricature from a single frontal image fast and automatically.
关键词:facial caricature;3D reconstruction;sparse morphable model;skin model
摘要:Frequent pattern mining has emerged as a key issue for analyzing the biological networks since it gives us insights into the organism and functional modules.A novel algorithm for this problem is proposed,which can efficiently obtain all these frequent subgraphs in networks based on the distribution of ring.To improve the accuracy of subgraph mining in non-exhaustive enumerate mode,additionally,we provide a dynamic sample algorithm.The experimental results in four real bio-networks show the superiority of our algorithm to existing algorithms.
摘要:In view of the gaps of current path coverage approach,this paper presents a new path coverage generation method based on the function call.According to control logic and function call syntax,all of the static function call paths are obtained.Dynamic path information is acquired after program running,and whether the test cases cover the program changes and the affected parts could be determined based on it.This method avoids the sharp increase in the number of paths,but also can guarantee that test completely.
摘要:Audio information hiding hides the secret message into the audio data flows by some characteristics of human auditory system.An audio information hiding algorithm with high-capacity which based on chaotic and wavelet transform is proposed,which,scrambles the secret message by chaotic sequences,and then one-to-one hides those into the high and low frequency of the wavelet field of the cover message by changing the cover message’s wavelet domain coefficient’s polarity based on the features that audio sample dots inversion has little influence on the human hearing.Experimental results show that this algorithm not only has good imperceptibility and is robust against different kinds of attacks,such as noise adding,re-quantizing,re-sampling,MP3 compression,and so on,but also has high-capacity,and can realize the blind detection.
SHEN Shi-qun, LIU Da-you, WANG Sheng-sheng, ZHU Li-na
Vol. 38, Issue 8, Pages: 1819-1824(2010)
摘要:Spatial database contains a great deal of topological and directional semantics.But traditional spatial data retrieval didn’t make good use of these high level semantics.To overcome this conceptual gap,this paper proposes a spatial data retrieval method based on sketch using 9-intersection model and Deep Direction-Relation Matrix.This method integrates direction relations and topological relations and can handle all data types in geographical databases.This paper outlines an algorithm based on the solutions of Binary CSP.Some methods to improve the efficiency of the algorithm for practical applications are also given.A prototype has been developed to experiment the method this paper proposed.
关键词:spatial data retrieval;direction relations;topological relations;spatial reasoning
摘要:An improved algorithm based on differential evolution algorithms,dynamic multi-group self-adaptive differential evolution algorithm with local search (DMSDELS),is applied to optimize functions in this paper.In DMSDELS,the population is randomly and dynamically divided into multi-group individuals,which can exchange information.To speed up search,in the mutation phase the best individual is chosen as the base vector,and the selection of the direction for difference vector is benefit to search.The scaling factor F and the crossover rate CR are self-adapted in order to balance the local search and the global search.To accelerate the convergence,elitist individuals could search in local after they explored specified generations.DMSDELS is tested on thirteen complex benchmark functions.The results are compared with those of other seven algorithms.The results show that DMSDELS is better in the search precision,convergence property and has strong ability to escape from the local sub-optima.
关键词:differential evolution algorithm;dynamic multi-group self-adaptive differential evolution algorithm with local search;optimization algorithm
摘要:At present,it is how to decrease the DNA volume that plays an important role in the development of DNA computing.In this paper,for the objective to reduce the DNA volume of 3-Dimensional Matching Problem which is a famous NP-complete problem,an improved DNA computing model based on the operations in Adleman-Lipton's model and the solution space of the sticker-based model is put forward.Then,the divide and comquer is introduced into the DNA computing and a new DNA computing's algorithm is proposed.In a computer simulation,the DNA strands of maximum number required was O(1.414n),the time complexity was O(15×n+30×q),the test tube complesity was O(1) and the longest DNA strand was O(15n/2+45q).Hence,the scale of 3-Dimensional Matching Problem may be enlarged from 67 to 134.This new algorithm is highly space-efficient and error-tolerant compared to the conventional brute-force searching,and can be scaled-up to solve large and hard 3-Dimensional Matching Problem.By the approach,we can also show DNA computing's vast potentials for resolving NP problems.
关键词:DNA computing;3-dimensional matching problem;divide and conquer;NP-complete problem
AI Shu-feng, YU Qun-ai, FENG Ji-ning, YANG Xiao-bo
Vol. 38, Issue 8, Pages: 1837-1840(2010)
摘要:Pseudo-satellites(pseudolie) is a kind of important technique in navigation and positioning,near-far effect is a problem that must be overcomed before using the pseudolie technique,a kind of receiver that can overcome near-far effect by using successive interference cancellation has been studied,the structure of the receiver is given, then the principle of determining the near-far effect and the key techniques of regenerating the pseudolie signal are given.Simulation results indicate that the receiver can be used in positing aided by pseudolie.
GE Yong-ming, ZHU Yi-hua, LONG Sheng-chun, PENG Jing
Vol. 38, Issue 8, Pages: 1841-1844(2010)
摘要:In an IEEE 802.11 based mobile ad hoc network (MANET),the wireless channel is accessed through the DCF (Distributed Coordinate Function),which is provided at the MAC (Medium Access Control) layer of the IEEE 802.11 standard to control channel contention among nodes.The DCF embeds the BEB (Binary Exponential Backoff) algorithm.The key parameter CW (contention window) of the BEB is investigated by stochastic analysis modeling so that the probability distribution of the contention window size is derived.Numeric results are presented to analyze the performance of the BEB.The derived probability distribution is applicable in MANETs.
关键词:IEEE 802.11 standard;mobile ad hoc networks;contention window;stochastic analysis
摘要:Aiming at the shortages of localization model in WSNs,a novel localization model based fuzzy recognition is advanced.In this model,a location space is divided into several small areas by some swatch nodes;every swatch node has its unique signal vector.An unknown node’s position can be calculated out by accounting the approach degree between swatch node’s signal vector and unknown node’s.This model adopts the RF signal as the foundation of localization;it can not only avoid the overlapped error in range-based localization model,but also can avoid the high complexity and the requirement of larger number reference nodes.The result of experiments in NS-2 shows that,this model has a high performance and is more suitable for applications in WSNs.
HUANG Chen, WANG Fu-rong, DAI Bin, YANG Jun, ZHANG Bo
Vol. 38, Issue 8, Pages: 1852-1857(2010)
摘要:Mobile stations in wireless Ad hoc network usually have limited energy and may prefer fewer transmissions for longer lifetime.We propose CRNC,an efficient data distribution mechanism based on network coding,that reduces data transmissions significantly,while increasing data transmission delay only slightly as compared to the protocol based on flooding.With extensive theoretical analysis and simulations,we show that CRNC based on network coding achieves a better tradeoff between resource usage and protocol performance.
XIAO Fang-ying, CHEN Han-wu, LIU Zhi-hao, LI Zhi-qiang, LIU Wen-jie
Vol. 38, Issue 8, Pages: 1858-1861(2010)
摘要:Cyclotomic cosets play very important roles in classical and quantum error correction theory.In order to constructing quantum BCH (Bose Chaudhuri Hocquenghem) codes with CSS constructing method from classical BCH codes,it needs to determine whether a BCH code contains its dual.It proposed several properties of cyclotomic cosets.And according to these properties,an algorithm with polynomial time complexity was presented to determine whether a non-primitive BCH code over finite field contains its dual code,which can also be applied to nonnarrow sense primitive BCH codes.
摘要:The shortcomings of increasing weight,interconnecting lines and cost exist in traditional metal housing technique for hermetic packaging of microwave multichip modules(MMCMs).Low Temperature Co-fired Ceramics(LTCC)cavity technique is an excellent hermetic packaging solution to realize integration of microwave interconnecting substrate and packaging house for developing MMCMs.The microwave LTCC cavity structures and their transitions were simulated and optimized by the microwave analysis software in this paper.Good agreement was acquired between the simulated results and the experimental results.A X-band receive channel in transmit/receive(T/R) module was developed successful by LTCC cavity technique.
关键词:low temperature co-fire ceramic;cavity structure;microwave multichip module;hermetic packaging
FAN Bin, ZHOU Tie-ge, YAN Shao-lin, YUE Hong-wei, WANG Zheng, SONG Feng-bin
Vol. 38, Issue 8, Pages: 1867-1891(2010)
摘要:To investigate the properties of the Josephson junction and its related circuits,a model of Josephson junction in Multisim has been created.Both the DC (Direct Current) Josephson effect and AC(Active Current) Josephson effect are testified using the model in Multisim.Subsequently,we observe a quasi-periodic motion in the Josephson junction by studying chaotic behaviors of the resistively-capacitively-inductively shunted junction.Besides,we investigate the characteristics of Shapiro steps in chaos and its special effect under thermal noises.Finally,rapid single flux quantum circuit is constructed to convert DC pulse to SFQ (Single Flux Quantum).The simulation results show that the model is powerful so that it can produce valuable predictions in the research of the Josephson junction,which plays an important role in analyzing and designing a superconductive device.
关键词:multisim;Josephson junction;chaotic behavior;shapiro steps;RSFQ(Rapid Single Flux Quantum)circuit
摘要:Small faults with insignificant abnormal symptoms were usually submerged in large faults or noise.Most fault diagnosis methods were invalid in the case when small faults occurred.We present a multi-level space decomposition mechanism and a small fault diagnosis algorithm.Implement designated component analysis (DCA) to the observation data for significant variation patterns;Remove the effect of significant designated patterns to get the residual which will increase the signal-to-noise rate of small fault signal;Determining whether small faults have occurred in the system using the projection significance index in the residual space;Repeat this process until all possible small faults are diagnosed.Simulation for observation data involved 4 faults shows its efficiency of this algorithm.
摘要:The rapid development of optical transmission and fast growth of Internet traffics require higher-performance routers.The overload of control plane often causes the oscillation of network.In a serious case,it results in router crashing.To overcome the problems of the monolithic control plane,some schemes of distributed control in a router have been proposed.In this paper,we thoroughly analyzed the problems of centralized control plane.Based on the critical obstacles in realizing router’s distributed control,we made summary and comparison of existing schemes from three aspects:distributed control plane,internal communication of distributed control plane and distributed implementation of routing protocol and algorithm.Finally,the future trend is discussed.
关键词:router;distributed control plane;internal communication;distributed routing protocol and algorithm
摘要:The framework of sensor management is presented.Its methods are summarized according to the level:sensor,platform,and network.The development of sensor management system is introduced.The status of sensor management research is reviewed.
关键词:sensor management;sensor scheduling;sensor allocation;sensor control
摘要:The current public key encryption schemes are vulnerable under the threat from quantum computing,it is necessary to research secure public key encryption algorithm under quantum computing.The concept of ergodic matrices and its property are given,and a new difficult problem named Two-side Ergodic Matrices Exponentiation (TEME) is proposed,which is proved at least NP-complete.Accordingly,we present a new public key encryption scheme based on ergodic matrices,and proved it is secure based on the TEME problem in the standard model,that is,the INDistinguishability against adaptively Chosen Ciphertext Attack (IND-CCA).
摘要:A novel sparsity adaptive subspace pursuit algorithm is proposed for sparse signals with unkown sparsity.First,a new sparsity estimation method based on matching test is used to get an initial estimated value.If subspace pursuit cannot reconstruct sparse signal successfully,the estimated value is increased each iteration.Weak matching is used in the iteration to select new atoms.Compared to other algorithms,it is competitive both in running speed and recovering accuracy.
摘要:Traditional Congestion control algorithms exhibit low convergence rate to equilibrium when the network capacity or round-trip delay is very large.In this paper,we propose a new algorithm called NMKC to accelerate the fairness convergence speed.To achieve this goal,our method incorporates two modes of operation based on the network feedback:low load mode updates the sending rate with MKC source control equation to achieve exponential convergence to efficiency,and high load mode constructs a novel source updating method to improve fairness convergence speed.The stability and convergence behavior is proved by theory.NS simulation results show that NMKC offers faster convergence speed.
关键词:congestion control;max-min Kelly control;fairness convergence;efficiency
WAN Chang-lin, HAN Xu, NIU Wen-jia, WANG Wen-jie, SHI Zhong-zhi
Vol. 38, Issue 8, Pages: 1923-1928(2010)
摘要:This paper proposes a semantic model for web service based on Dynamic Description Logic (DDL).In this paper,service composition is divided into two stages:logic plan stage and grounding stage.Therefore,a service composition problem is reduced to an action planning problem at the planning stage and a QoS-aware service selection problem at the grounding stage.This paper mainly focuses on the QoS model and service selection.A composite service is divided to several smaller services,which are then solved separately by a recursive branch-and-bound algorithm.Experiments show that the proposed approach is efficient and scalable for runtime service selection,and better handles complicated structure of service composition than existing approaches.
摘要:The unreasonable allocation scheme of bandwidth resources is a serious problem in the opening Internet.To restrict the users’ selfishness,a network bandwidth allocation mechanism based on VCG(Vickrey-Clarke-Groves) auction is proposed,which has the dominated strategy incentive compatible property.A discrete stochastic learning algorithm which is used to guide the users to choose strategies is introduced and its convergence is analyzed further.Simulation results show that the selfish users do not have any incentives to lie and provide the real bandwidth requirements,through the effective punishment scheme;the discrete stochastic learning algorithm can make the users select the dominate strategy correctly and allocate the bandwidth resources reasonably.
XU Xiang-li, ZHANG Li-biao, LIU Xiang-dong, YU Zhe-zhou, ZHOU Chun-guang
Vol. 38, Issue 8, Pages: 1935-1940(2010)
摘要:The evolution searching process of particle swarm optimization(PSO) algorithm and user feedback process are combined effectively,and an image retrieval relevance feedback(RF) algorithm based on PSO is proposed,in which the influence to user perception and limitation to feedback effect caused by initial retrieval are avoided.The algorithm makes users'understanding to the retrieval goal become clear gradually,could search image database effectively and comprehensively,and at the same time it could avoid the contradiction of efficiency and retrieval effect caused by multi-feedback.Experiments have verified the validity of the proposed algorithm.
LI You-ming, SHEN Wei, ZHANG Jing-yan, XU Tie-feng
Vol. 38, Issue 8, Pages: 1941-1944(2010)
摘要:Crosstalk is the major limiting issue in VDSL2 system.A class of low complexity and effective iterative crosstalk cancellation algorithms based on matrix splitting and iterative theory are proposed for VDSL2 upstream transmission.Two special iterative algorithms based on diagonal matrix splitting and up triangular matrix splitting are discussed,as the two splitting inspired from using the channel matrix property of diagonal dominant and column wise diagonal dominant respectively.These results reveal that the iterative method based on up triangular matrix splitting is superior to other iterative algorithms,and its performance approaches zero forcing algorithm,but the complexity in each iteration step is almost the same as first order algorithm.Computer simulation results based on measured data verify the efficiency of the new proposed method.
摘要:A PDH to 622 Mb/s SDH/SONET mapping & de-mapping chip is proposed in this paper,which integrates DS1/E1/J1 framer,DS1/DS3 multiplexer and E1/E3 multiplexer.It provides high speed interface for 622 Mb/s and 155 Mb/s SDH/SONET signals and three add-drop-multiplexer bus interfaces for STM-1/STS-3 signals.It supports both MSP 1+1 protection switching and ring (UPSR) network topologies.A single chip can provide complete STM-1/STS-3 mapping & de-mapping for up to 84 DS1/ J1 or 63 E1 signals,and STM-4/STS-12 mapping & de-mapping for several DS3/E3/STS-1 signals.Point-to-point and ring applications are supported,and in switching mode 2016 DS0/E0 application is supported.Four chips can implement complete STM-4/STS-12 mapping de-mapping for 336 DS1/J1 or 252 E1 tributary signals.The chip is fabricated in TSMC 0.13 μm CMOS process and the gate count is about 6 million.It is packaged with 700-pin PGBA technology and meets optical network transmission requirement.It has been successfully used in optical network communication devices.
MIAO Duo-qian, CHEN Yu-ming, WANG rui-zhi, ZHANG Hong-yun
Vol. 38, Issue 8, Pages: 1952-1957(2010)
摘要:Knowledge reduction is widely studied under algebra view and information view.In this paper,knowledge reduction under graph view is presented.A complete recursive algorithm for minimal reduction under graph view is designed.In virtue of graph searching methods of artificial intelligence,rotation pruning operator and backtracking pruning operator for answering the minimal reduction question are proposed.These methods’ completeness for the minimal reduction is proved.In order to test the efficiency of the algorithm,some experiments are made on simulative data.Theoretical analysis and experimental results show that the reduction algorithm under graph view is efficient and feasible.
LIU Chun-yang, TAN Ying-qing, LIU Chang-an, MA Ying-wei
Vol. 38, Issue 8, Pages: 1958-1962(2010)
摘要:A multi-agent reinforcement learning method based on voting to solve the collaboration problem of team members is presented.The method translates the collaboration problem into learning strategies by defining joint actions which called the strategies and then can simplify the problem.Through dividing of the playground,the location can be measured by a lot of numbered regions and then can effectively reduce the state-space dimensions to speed up the pace of learning.By distinguishing the environment states and taking the collaboration status into account,that causing the reduction of the state-action space,the learning speed can be further improved.Using a voting process that combines the decisions of the agents can realize the collaboration.At last,experimental results show the effectiveness and correctness of the method.
摘要:In order to realize a simple topology,high efficiency,high frequency,low voltage stress,easy to control soft switching three phase inverter,A novel Passive Assisted Soft Switching Converter is presented in the paper.Soft Switching of the switch can be achieved by using Passive Assisted Resonant network.It is very attractive for high power application where IGBT is predominantly used as the power switch.Its operation principle is analyzed through its application to the boost converter.The condition of soft switching and the design considerations are analyzed in detail.The novel soft switching cell can be also used in other basic dc-dc converter.A 5kW~20kHz prototype which uses IGBT (insulated gate bipolar transistor) is made.The effectiveness of the proposed converter is confirmed by the experimental results.
关键词:passive DC-DC converter;zero current switching (ZCS);zero voltage switching (ZVS)
KONG Wan-zeng, SUN Zhi-hai, YANG Can, DAI Guo-jun, SUN Changsihe
Vol. 38, Issue 8, Pages: 1980-1985(2010)
摘要:To deal with the problem that classical spectral clustering methods can not automatically determine the number of class.A new algorithm called automatic spectral clustering(ASC) based on eigengap and orthogonal eigenvector was presented in this paper.The proposed method first constructed the affinity matrix of data,and gained series of eigenvalues and eigenvectors through spectral decomposition.Second,ordered the eigenvalues and used the first maximum eigengap to determine the number of classes.The data was classified by the class number and the angle between two eigenvectors as similarity.The effectiveness of the proposed algorithm was verified on artificial data,and was compared with k-means,FCM and Jordan algorithm on UCI database.The experiment results demonstrate that the proposed method ASC outperforms other three methods in respect of classification accuracy.