In practice, retraining a trained classifier is necessary when novel data become available. This paper adopts an incremental learning procedure to adaptively train a Kernel-based Nonlinear Representor (KNR), a recentl...In practice, retraining a trained classifier is necessary when novel data become available. This paper adopts an incremental learning procedure to adaptively train a Kernel-based Nonlinear Representor (KNR), a recently presented nonlinear classifier for optimal pattern representation, so that its generalization ability may be evaluated in time-variant situation and a sparser representation is obtained for computationally intensive tasks. The addressed techniques are applied to handwritten digit classification to illustrate the feasibility for pattern recognition.展开更多
The paper presents a new algorithm of NonLinearly Adaptive Interpolation (NLAI). NLAI is based on both the gradients and the curvature of the signals with the predicted subsection. It is characterized by adap- tive no...The paper presents a new algorithm of NonLinearly Adaptive Interpolation (NLAI). NLAI is based on both the gradients and the curvature of the signals with the predicted subsection. It is characterized by adap- tive nonlinear interpolation method with extracting the characteristics of signals. Experimental research testi- fies the validity of the algorithm using the echoes of the Ground Penetrating Radar (GPR). A comparison of this algorithm with other traditional algorithms demonstrates that it is feasible.展开更多
Integrating with practical e-commerce application, this paper introduces a novel multi-dimension evaluation method to depict and calculate the trust values. The multi-dimension evaluation metrics include functional an...Integrating with practical e-commerce application, this paper introduces a novel multi-dimension evaluation method to depict and calculate the trust values. The multi-dimension evaluation metrics include functional and nonfunctional properties and corresponding weights. The continuous measurement values and the Markov chain mechanism are adopted to compute the trust value and detect the malicious behaviors. The current evaluation has larger influence factor on the next transaction behavior. A trust model is implemented with web service which consists of publication, filtrating, calculating and storage center. It is easily extended and the user only defines each property and its weights according to specific requirements, then the trust values are got. In order to conveniently manage and avoid the dead-lock, some constraint rules are proposed. The results show that the method based on multi-dimension can reflect objectively the dynamic change of trust values.展开更多
基金Supported by the Key Project of Chinese Ministry of Education (No.105150).
文摘In practice, retraining a trained classifier is necessary when novel data become available. This paper adopts an incremental learning procedure to adaptively train a Kernel-based Nonlinear Representor (KNR), a recently presented nonlinear classifier for optimal pattern representation, so that its generalization ability may be evaluated in time-variant situation and a sparser representation is obtained for computationally intensive tasks. The addressed techniques are applied to handwritten digit classification to illustrate the feasibility for pattern recognition.
基金Supported by the National Natural Science Foundation of China (No.60572152).
文摘The paper presents a new algorithm of NonLinearly Adaptive Interpolation (NLAI). NLAI is based on both the gradients and the curvature of the signals with the predicted subsection. It is characterized by adap- tive nonlinear interpolation method with extracting the characteristics of signals. Experimental research testi- fies the validity of the algorithm using the echoes of the Ground Penetrating Radar (GPR). A comparison of this algorithm with other traditional algorithms demonstrates that it is feasible.
文摘Integrating with practical e-commerce application, this paper introduces a novel multi-dimension evaluation method to depict and calculate the trust values. The multi-dimension evaluation metrics include functional and nonfunctional properties and corresponding weights. The continuous measurement values and the Markov chain mechanism are adopted to compute the trust value and detect the malicious behaviors. The current evaluation has larger influence factor on the next transaction behavior. A trust model is implemented with web service which consists of publication, filtrating, calculating and storage center. It is easily extended and the user only defines each property and its weights according to specific requirements, then the trust values are got. In order to conveniently manage and avoid the dead-lock, some constraint rules are proposed. The results show that the method based on multi-dimension can reflect objectively the dynamic change of trust values.
