An effective ensemble should consist of a set of networks that are both accurate and diverse. We propose a novel clustering-based selective algorithm for constructing neural network ensemble, where clustering technolo...An effective ensemble should consist of a set of networks that are both accurate and diverse. We propose a novel clustering-based selective algorithm for constructing neural network ensemble, where clustering technology is used to classify trained networks according to similarity and optimally select the most accurate individual network from each cluster to make up the ensemble. Empirical studies on regression of four typical datasets showed that this approach yields significantly smaller en- semble achieving better performance than other traditional ones such as Bagging and Boosting. The bias variance decomposition of the predictive error shows that the success of the proposed approach may lie in its properly tuning the bias/variance trade-off to reduce the prediction error (the sum of bias2 and variance).展开更多
According to the features of movements of humanoid robot, a control system for humanoid robot walking on uneven terrain is present. Constraints of stepping over stairs are analyzed and the trajectories of feet are cal...According to the features of movements of humanoid robot, a control system for humanoid robot walking on uneven terrain is present. Constraints of stepping over stairs are analyzed and the trajectories of feet are calculated by intelligent computing methods. To overcome the shortcomings resulted from directly controlling the robot by neural network （NN） and fuzzy logic controller （FLC）, a revised particle swarm optimization （PSO） algorithm is proposed to train the weights of NN and rules of FLC. Simulations and experiments on different control methods are achieved for a detailed comparison. The results show that using the proposed methods can obtain better control effect.展开更多
Converting between “fuzzy concept” and “numerical value” in computer aided assessment is rather difficult in many applications. This paper presents a LVQ neural network paradigm for sensory evaluation. This intell...Converting between “fuzzy concept” and “numerical value” in computer aided assessment is rather difficult in many applications. This paper presents a LVQ neural network paradigm for sensory evaluation. This intelligent approach utilizes predefined class information for supervised learning in order to solve the converting problem and keep the fuzziness and imprecision of the whole sensory information. The method is validated by the experiment on stimulation evaluation of cigarette sensory.展开更多
This paper focuses mainly on application of Partially Connected Backpropagation Neural Network (PCBP) instead of typical Fully Connected Neural Network (FCBP). The initial neural network is fully connected, after trai...This paper focuses mainly on application of Partially Connected Backpropagation Neural Network (PCBP) instead of typical Fully Connected Neural Network (FCBP). The initial neural network is fully connected, after training with sample data using cross-entropy as error function, a clustering method is employed to cluster weights between inputs to hidden layer and from hidden to output layer, and connections that are relatively unnecessary are deleted, thus the initial network becomes a PCBP network. Then PCBP can be used in prediction or data mining by training PCBP with data that comes from database. At the end of this paper, several experiments are conducted to illustrate the effects of PCBP using Iris data set.展开更多
To increase predictive behaviors of neural network dynamic model, an experimental case study of a new approach to systems controller design is presented. The experiment is based on neural networks inverse plant model....To increase predictive behaviors of neural network dynamic model, an experimental case study of a new approach to systems controller design is presented. The experiment is based on neural networks inverse plant model. Special rules for network training are developed. Such system is close to model-based predictive control, but needs much less computational resources. The approach advantages are shown by the control of laboratory complex plants.展开更多
This paper briefly introduces the collection and recognition of bio-medical sig nals, designs the method to collect FM signals. A detailed discussion on the sys tem hardware, structure and functions is also given. Und...This paper briefly introduces the collection and recognition of bio-medical sig nals, designs the method to collect FM signals. A detailed discussion on the sys tem hardware, structure and functions is also given. Under LabWindows/CVI,the ha rdware and the driver do compatible, the hardware equipment work properly active ly. The paper adopts multi threading technology for real-time analysis and make s use of latency time of CPU effectively, expedites program reflect speed, impro ve s the program to perform efficiency. One threading is collecting data; the other threading is analyzing data. Using the method, it is broaden to analyze the sig nal in real-time. Wavelet transform to remove the main interference in the FM a nd by adding time-window to recognize with BP network; Finally the results of c ollecting signals and BP networks are discussed.8 pregnant women’s signals of F M were collected successfully by using the sensor. The correct of BP network rec ognition is about 83.3% by using the above measure.展开更多
文摘An effective ensemble should consist of a set of networks that are both accurate and diverse. We propose a novel clustering-based selective algorithm for constructing neural network ensemble, where clustering technology is used to classify trained networks according to similarity and optimally select the most accurate individual network from each cluster to make up the ensemble. Empirical studies on regression of four typical datasets showed that this approach yields significantly smaller en- semble achieving better performance than other traditional ones such as Bagging and Boosting. The bias variance decomposition of the predictive error shows that the success of the proposed approach may lie in its properly tuning the bias/variance trade-off to reduce the prediction error (the sum of bias2 and variance).
基金This material is based upon work funded by State Key Laboratory of Robotics and System （HIT） Foundation of China under Grant No. SKLRS-2012-MS-06, China Postdoctoral Science Foundation under Grant No. 2013M531022, Research project of laboratory work in universities of Zhejiang Province under Grant No. ZD201504, Educational technology research program of Zhejiang Province under Grant No. JA027.
文摘According to the features of movements of humanoid robot, a control system for humanoid robot walking on uneven terrain is present. Constraints of stepping over stairs are analyzed and the trajectories of feet are calculated by intelligent computing methods. To overcome the shortcomings resulted from directly controlling the robot by neural network （NN） and fuzzy logic controller （FLC）, a revised particle swarm optimization （PSO） algorithm is proposed to train the weights of NN and rules of FLC. Simulations and experiments on different control methods are achieved for a detailed comparison. The results show that using the proposed methods can obtain better control effect.
文摘Converting between “fuzzy concept” and “numerical value” in computer aided assessment is rather difficult in many applications. This paper presents a LVQ neural network paradigm for sensory evaluation. This intelligent approach utilizes predefined class information for supervised learning in order to solve the converting problem and keep the fuzziness and imprecision of the whole sensory information. The method is validated by the experiment on stimulation evaluation of cigarette sensory.
文摘This paper focuses mainly on application of Partially Connected Backpropagation Neural Network (PCBP) instead of typical Fully Connected Neural Network (FCBP). The initial neural network is fully connected, after training with sample data using cross-entropy as error function, a clustering method is employed to cluster weights between inputs to hidden layer and from hidden to output layer, and connections that are relatively unnecessary are deleted, thus the initial network becomes a PCBP network. Then PCBP can be used in prediction or data mining by training PCBP with data that comes from database. At the end of this paper, several experiments are conducted to illustrate the effects of PCBP using Iris data set.
文摘To increase predictive behaviors of neural network dynamic model, an experimental case study of a new approach to systems controller design is presented. The experiment is based on neural networks inverse plant model. Special rules for network training are developed. Such system is close to model-based predictive control, but needs much less computational resources. The approach advantages are shown by the control of laboratory complex plants.
文摘This paper briefly introduces the collection and recognition of bio-medical sig nals, designs the method to collect FM signals. A detailed discussion on the sys tem hardware, structure and functions is also given. Under LabWindows/CVI,the ha rdware and the driver do compatible, the hardware equipment work properly active ly. The paper adopts multi threading technology for real-time analysis and make s use of latency time of CPU effectively, expedites program reflect speed, impro ve s the program to perform efficiency. One threading is collecting data; the other threading is analyzing data. Using the method, it is broaden to analyze the sig nal in real-time. Wavelet transform to remove the main interference in the FM a nd by adding time-window to recognize with BP network; Finally the results of c ollecting signals and BP networks are discussed.8 pregnant women’s signals of F M were collected successfully by using the sensor. The correct of BP network rec ognition is about 83.3% by using the above measure.
