Aiming at the problem of privacy leakage in the application of traditional clustering algorithm, a spectral clustering algorithm based on differential privacy protection was proposed. Based on the differential privacy model, the cumulative distribution function was used to generate random noise that satisfies Laplasse distribution. Then the noise was added to the sample similarity function calculated by the spectral clustering algorithm, which disturbed the weight values between the individual samples and realized information hiding between sample individuals for privacy protection. Experimental results of UCI dataset verify that the proposed algorithm can achieve effective data clustering within a certain degree of information loss, and can also protect clustered data.

With the development of bioinformatics, gene expression microarray and image recognition, classification on high-dimensional and small-sample-size data has become a challenging task in data ming, machine learning and pattern recognition as well. High-dimensional and small-sample-size data may cause the problem of "curse of dimensionality" and overfitting. Feature selection can prevent the "curse of dimensionality" effectively and promote the generalization ability of classification mode, and thus become a hot research topic. Accordingly, some recent development of world-wide research on feature selection in high-dimensional and small-sample-size classification was briefly reviewed. Firstly, the nature of high-dimensional and small-sample feature selection was analyzed. Secondly, according to their essential difference, feature selection algorithms for high-dimensional and small-sample-size classification were divided into four categories and compared to summarize their advantages and disadvantages. Finally, challenges and prospects for future trends of feature selection in high-dimensional small-sample-size data were proposed.

Focused on the issue that large-scale network has characteristics of huge number of nodes, high structural complexity and difficulty to demonstrate its structural characteristics by the limited screen space, a multi-level network visualization method based on community detection was proposed. Firstly, a community detection algorithm based on network modularity was used to detect the network node and a greedy algorithm was used to find the community detection with maximum modularity to get different level of granularity communities. Then, in order to solve the problem that the Force-Directed Algorithm (FDA) could not display network nodes hierarchically, the classic FDA was improved by setting the level blinding force to achieve hierarchical layout of different level of granularity communities. Finally, high level communities and low level nodes were displayed respectively by using the interactive method such as multi-window view and Overview+Detail, meeting the requirement of both network high-level macrostructure and low-level details of the display. In the simulation test, the community detection algorithm is faster and more accurate compared to self-contained GN (Girvan-Newman) algorithm. The theoretical analysis and simulation results show that the proposed method has good effect and performance in display and interaction of large-scale network structure.

In order to improve the security of secure communication, a new Generalized Hybrid Dislocated Function Projective Synchronization (GHDFPS) based on generalized hybrid dislocated projective synchronization and function projective synchronization was researched by Lyapunov stability theory and adaptive active control method. At the same time, the control methods of GHDFPS between two different-order chaotic systems with uncertain parameter and parameter identification were presented, and the application of the novel synchronization on secure communication was analyzed. By strict mathematical proof and numerical simulation, the GHDFPS between two different-order chaotic systems with uncertain parameter were achieved, the uncertain parameter was identified. Because of the variety of function scaling factor matrix, the security of secure communication has been increased by GHDFPS. Moreover, this synchronization form and method of control were applied to secure communication via chaotic masking modulation. Many information signals can be recovered and validated.

Accurate extraction of a target’s edge in a log-polar space is a precondition and key point to successfully apply the visual invariance of the log-polar transformation. Since it is impossible for traditional algorithms to extract the single-pixel edge in an intensively noisy environment, a unique edge extraction algorithm on the basis of active contour model and level set method was designed. After noise removal on the whole via Canny operator based level set method, the energy-driving active contour model was used to iteratively approach the potential edges. By clearing out false edges with an improved tracing way, the true target’s edge was extracted finally. The experimental results demonstrate the effective performance of the proposed algorithm with the edge feature similarity up to 96%.

In order to improve the ability of Artificial Bee Colony (ABC) algorithm at exploitation, a new Chaos Artificial Bee Colony (CH-ABC) algorithm was proposed for continuous function optimization problems. A new chaotic local search operator was embedded in the framework of the new algorithm. The new operator, whose search radius shrinks with the evolution generation, can do the local search around the best food source. The simulation results show that: compared with those of ABC algorithm, the solution quality and the convergence speed of the new algorithm are better for Rosenbrock and the convergence speed of the new algorithm is better for Griewank and Rastrigin.

In order to resolve the widely existing problem of Non-Transitive Connectivity (NTC) in networks, a uncertainty Distributed Hash Tables （DHT） method to resolve the NTC problem in overlay network was proposed. The relationship of bottom node ID and logical space location was lifted to avoid infection of network structure by NTC nodes, and redirection route mechanism was used to implement the message routing of network. Simulation shows that uncertainty DHT could effectively resolve the NTC problem in overlay network.

An artificial neural net(ANN) model was established for considering multi physical factors’ effects on reading environment.Typical back-propagation net(BP net) with three layers was selected with input parameters of illuminance and desktop’s brightness,outputing relative reading gross and reading accuracy.The ANN model could consider several factors’ effects on reading working at the same time,figuring out multi-parameters and strong coupling in reading environment.Using the ANN model,illuminance and brightness were analyzed,and relative reading gross and reading accuracy were calculated under different reading environments.

Based on the theory of kernel generalized optimal feature extracted mode,a new method for the corresponding mode was proposed.Firstly space transform method was used to transform initial kernel between class scatter matrix and kernel total scatter matrix,so the kernel total scatter matrix became positive definition. At the same time,by the means of kernel uncorrelated feature vectors extraction,the feature vectors got were statistical uncorrelated.To verify the effectiveness of this method,experiment was tested on ORL face databases and the result showed that the face recognition method proposed is more available than other methods such as kernel discriminant analysis.