Yunze Cai

A hybrid genetic algorithm for feature selection wrapper based on mutual information

In this study, a hybrid genetic algorithm is adopted to find a subset of features that are most relevant to the classification task. Two stages of optimization are involved. The outer optimization stage completes the global search for the best subset of features in a wrapper way, in which the mutual information between the predictive labels of a trained classifier and the true classes serves as the fitness function for the genetic algorithm. The inner optimization performs the local search in a filter manner, in which an improved estimation of the conditional mutual information acts as an independent measure for feature ranking taking account of not only the relevance of the candidate feature to the output classes but also the redundancy to the already-selected features. The inner and outer optimizations cooperate with each other and achieve the high global predictive accuracy as well as the high local search efficiency. Experimental results demonstrate both parsimonious feature selection and excellent classification accuracy of the method on a range of benchmark data sets.

A new image fusion performance metric based on visual information fidelity

Because subjective evaluation is not adequate for assessing work in an automatic system, using an objective image fusion performance metric is a common approach to evaluate the quality of different fusion schemes. In this paper, a multi-resolution image fusion metric using visual information fidelity (VIF) is presented to assess fusion performance objectively. This method has four stages: (1) Source and fused images are filtered and divided into blocks. (2) Visual information is evaluated with and without distortion information in each block. (3) The visual information fidelity for fusion (VIFF) of each sub-band is calculated. (4) The overall quality measure is determined by weighting the VIFF of each sub-band. In our experiment, the proposed fusion assessment method is compared with several existing fusion metrics using the subjective test dataset provided by Petrovic. We found that VIFF performs better in terms of both human perception matching and computational complexity.

A reduced-order approach to filtering for systems with linear equality constraints

In this paper, the filtering problem is investigated for a class of discrete systems with linear equality constraints. The system under consideration is subject to both noises and time-varying constrained conditions. Attention is focused on the design of a new reduced-order filter under a mild assumption such that the estimation performance of the proposed filter outperforms those of the traditional filters. By using the reorganized constraint information, the original system is transformed to a reduced-order system. A new recursive state estimator is developed, which is proved to have higher estimation precision than several existing filters. Subsequently, further analysis shows that the constrained Kalman predictor is a special case of the proposed filter. Finally, a numerical example is employed to demonstrate the effectiveness of our approach.

Evolution of cooperation among mobile agents

We study the effects of mobility on the evolution of cooperation among mobile players, which imitate collective motion of biological flocks and interact with neighbors within a prescribed radius R. Adopting the the prisoner’s dilemma game and the snowdrift game as metaphors, we find that cooperation can be maintained and even enhanced for low velocities and small payoff parameters, when compared with the case that all agents do not move. But such enhancement of cooperation is largely determined by the value of R, and for modest values of R, there is an optimal value of velocity to induce the maximum cooperation level. Besides, we find that intermediate values of R or initial population densities are most favorable for cooperation, when the velocity is fixed. Depending on the payoff parameters, the system can reach an absorbing state of cooperation when the snowdrift game is played. Our findings may help understanding the relations between individual mobility and cooperative behavior in social systems.

A Wrapper for Feature Selection Based on Mutual Information

This paper adopts a wrapper method to find a subset of features that are most relevant to the classification task. The approach utilizes an improved estimation of the conditional mutual information which is used as an independent measure for feature ranking in the local search operations. Meanwhile, the mutual information between the predictive labels of a trained classifier and the true classes is used as the fitness function in the global search for the best subset of features. Thus, the local and global searches consist of a hybrid genetic algorithm for feature selection. Experimental results demonstrate both parsimonious feature selection and excellent classification accuracy of the method on a range of benchmark data sets

Consensus tracking for multi-agent systems with directed graph via distributed adaptive protocol

The consensus tracking problem is investigated for multi-agent systems with directed graph. To avoid using any global information, a novel adaptive protocol is proposed based only on the relative state information. A monotonically increasing function for each agent is inserted into the protocol to provide extra freedom for design. By using matrix theory and appropriate Lyapunov techniques, it is shown that the consensus tracking can be achieved in a fully distributed fashion if agents? dynamics are stabilizable and the topological graph contains a directed spanning tree with the leader as the root node. A simulation example shows the effectiveness of the design method.

An empirical comparison of ensemble classification algorithms with support vector machines

An ensemble classifier often has better performance than any of the single learned classifiers in the ensemble. In this paper, the trained support vector machine (SVM) classifiers are used as basic classifiers. The ensemble methods for creating ensemble classifier, such as bagging and boosting, etc., are evaluated on two data sets. Some conclusions are obtained. Bagging with SVM can stably improve classification accuracy, while the improvement obtained by boosting with SVM is not obvious. These two methods largely increase space complexity and time complexity. Comparatively, the multiple SVM decision model, training individual SVM classifiers using training subsets obtained by partitioning the original training set, has a better trade-off between the classification accuracy and efficiency.

Evolutionary prisoner’s dilemma game in flocks

We investigate an evolutionary prisoner’s dilemma game among self-driven agents, where collective motion of biological flocks is imitated through averaging directions of neighbors. Depending on the temptation to defect and the velocity at which agents move, we find that cooperation can not only be maintained in such a system but there exists an optimal size of interaction neighborhood, which can induce the maximum cooperation level. When compared with the case that all agents do not move, cooperation can even be enhanced by the mobility of individuals, provided that the velocity and the size of neighborhood are not too large. Besides, we find that the system exhibits aggregation behavior, and cooperators may coexist with defectors at equilibrium.

A Filter Approach to Feature Selection Based on Mutual Information

In pattern recognition, feature selection aims to choose the smallest subset of features that is necessary and sufficient to describe the target concept. In this paper, a mutual information-based constructive criterion under arbitrary information distributions of input features is presented for feature selection. This criterion can capture both the relevance to the output classes and the redundancy with respect to the already-selected features without any parameters like beta in MIFS or MIFS-U methods to be preset. Furthermore, a modified greedy feature selection algorithm called MICC is proposed, and experimental results demonstrate the good performance of MICC on both synthetic and benchmark data sets

Rough sets method for SVM data preprocessing

To improve the generalization performance and structure of SVM classifiers (SVCs), we introduce rough sets theory to the data preprocessing of SVCs. Three measures are taken: removing duplicate samples from the dataset, finding a reduct and then multiplying every attribute with its corresponding significance factor which equals to the dependency of decision attribute with respect to the attribute. Experiment results on a UCI benchmark dataset and a practical steam turbine failure diagnosis problem show that the presented approach is feasible