Beijing Key Laboratory of Computational Intelligence and Intelligent System

Beijing, China

Beijing Key Laboratory of Computational Intelligence and Intelligent System

Beijing, China
SEARCH FILTERS
Time filter
Source Type

Wu Z.,Beijing University of Technology | Wang L.,Beijing University of Technology | Wang L.,Beijing Key Laboratory of Computational Intelligence and Intelligent System
Proceedings of the 29th Chinese Control and Decision Conference, CCDC 2017 | Year: 2017

Most man-made indoor and urban scenes are composed of a set of orthogonal and parallel planes. In robotics and computer vision, these scenes typically represented by the Manhattan-World model. The accurate estimation of the Manhattan Frame, which consists of three orthogonal directions being used to represent the Manhattan-World, plays an important role in many applications, such as SLAM, scene understanding and 3D reconstruction. In this paper, a new method for accurately recovering the Manhattan frame from a single RGB-D image by using the orientation relevance is proposed. It first extracts planes from the input single RGB-D image. Then three orthogonal dominant planes are determined by introducing the concept of orientation relevance. Finally, the Manhattan Frame can be easily recovered from the obtained three orthogonal dominant planes. Experiments with open dataset validate the proposed method. The overall performance of the proposed method, which takes both accuracy and speed into account, is superior to that of the state-of-the-art methods. It is also applied on the application of scene annotation to confirm its applicability. © 2017 IEEE.


Han H.-G.,Beijing University of Technology | Han H.-G.,Beijing Key Laboratory of Computational Intelligence and Intelligent System | Lin Z.-L.,Beijing University of Technology | Lin Z.-L.,Beijing Key Laboratory of Computational Intelligence and Intelligent System | And 2 more authors.
Neurocomputing | Year: 2017

In this paper, a self-organizing fuzzy neural network with adaptive gradient algorithm (SOFNN-AGA) is proposed for nonlinear systems modeling. First, a potentiality of fuzzy rules (PFR) method is introduced by using the output of normalized layer and the error reduction ratio (ERR) in the training process. And a structure learning approach is developed to determine the network size based on PFR. Second, a novel adaptive gradient algorithm (AGA) with adaptive learning rate is designed to adjust the parameters of SOFNN-AGA. Moreover, a theoretical analysis on the convergence of SOFNN-AGA is given to show the efficiency in both fixed structure and self-organizing structure cases. Finally, to demonstrate the merits of SOFNN-AGA, simulation and experimental results of several benchmark problems and a real world application are examined for nonlinear systems modeling with comparisons against other existing methods. Some promising results are reported in this study, indicating that the proposed SOFNN-AGA performs better favorably in terms of both convergence speed and modeling accuracy. © 2017.


Li W.,Beijing University of Technology | Li W.,Beijing Key Laboratory of Computational Intelligence and Intelligent System | Qiao J.,Beijing University of Technology | Qiao J.,Beijing Key Laboratory of Computational Intelligence and Intelligent System
Beijing Gongye Daxue Xuebao/Journal of Beijing University of Technology | Year: 2017

Facing the structure design problem of fuzzy neural networks (FNNs), this paper proposed a structure design approach based on the recursive clustering and similarity methods. First, a recursive clustering method to identify FNN structure was proposed. Guided by the strength of output variations and using the recursive sub-clustering as the means, the proposed method determined the initial network structure through recursive iterations. Second, maintaining a high accuracy, the method calculated the similarity degree between each pair of fuzzy rules and then merged highly similar rules to simplify the initialized structure of the FNN. Finally, numerical experiments in function approximation and nonlinear system identification were used to verify the feasibility and effectiveness of the proposed approach. © 2017, Editorial Department of Journal of Beijing University of Technology. All right reserved.


Qiao J.-F.,Beijing University of Technology | Qiao J.-F.,Beijing Key Laboratory of Computational Intelligence and Intelligent System | Zhou H.-B.,Beijing University of Technology | Zhou H.-B.,Beijing Key Laboratory of Computational Intelligence and Intelligent System | Zhou H.-B.,Huaiyin Institute of Technology
Kongzhi Lilun Yu Yingyong/Control Theory and Applications | Year: 2017

A novel online self-organizing fuzzy neural network (FNN) based on the improved Levenberg-Marquardt (ILM) learning algorithm and singular value decomposition (SVD) is proposed to predict the effluent total phosphorus (TP) in a wastewater treatment process. The centers and widths of membership functions and weights of output layer are trained by ILM learning algorithm. Meanwhile, the output matrix of the rule layer is decomposed with SVD, which is implemented by one-sided Jacobi's transformation. The neurons of rule layer are adjusted dynamically with growing and pruning algorithms, which are based on the singular values. In addition, the convergence of the proposed ILM--SVDFNN has been proved both in the structure fixed phase and the structure adjusting phase. Finally, the validity and practicability of the model are illustrated with three examples, including typical nonlinear system identification, Mackey-Glass time series prediction, and prediction of effluent TP. Simulation results demonstrate that the proposed ILM--SVDFNN generates a fuzzy neural network automatically and effectively with a highly accurate and compact structure, and it can well satisfy the detection accuracy and real-time requirements of the prediction of effluent TP. ©2017, Editorial Department of Control Theory & Applications South China University of Technology. All right reserved.


Yan A.,Beijing University of Technology | Yan A.,Beijing Key Laboratory of Computational Intelligence and Intelligent System | Wang D.,La Trobe University
Expert Systems with Applications | Year: 2015

To achieve better classification performance using case-based reasoning classifiers, we propose a retrieval-based revision method with trustworthiness evaluation for problem solving. An improved case evaluation method is employed to evaluate the trustworthiness of the suggested solution after the reuse step, which will divide the target cases and its suggested solutions into a trustworthy set and an untrustworthy set in accordance with a threshold value of trustworthiness. The attribute weights are adjusted by running a genetic algorithm and are used in the second round of retrieval of the untrustworthy set to obtain the classification results. Experimental results demonstrate that our proposed method performs favorably compared with other methods. Also, the proposed method has less computation complexity for the trustworthiness evaluation, and enhances understanding on thinking and inference for case-based reasoning classifiers. © 2015 Elsevier Ltd. All rights reserved.


Li W.,Beijing University of Technology | Li W.,CAS Institute of Automation | Li W.,Beijing Key Laboratory of Computational Intelligence and Intelligent System | Li J.,Capital Medical University | And 5 more authors.
Neural Plasticity | Year: 2016

Previous neuroimaging studies suggested structural or functional brain reorganizations occurred in prelingually deaf subjects. However, little is known about the reorganizations of brain network architectures in prelingually deaf adolescents. The present study aims to investigate alterations of whole-brain functional network using resting-state fMRI and graph theory analysis. We recruited 16 prelingually deaf adolescents (1018 years) and 16 normal controls matched in age and gender. Brain networks were constructed from mean time courses of 90 regions. Widely distributed network was observed in deaf subjects, with increased connectivity between the limbic system and regions involved in visual and language processing, suggesting reinforcement of the processing for the visual and verbal information in deaf adolescents. Decreased connectivity was detected between the visual regions and language regions possibly due to inferior reading or speaking skills in deaf subjects. Using graph theory analysis, we demonstrated small-worldness property did not change in prelingually deaf adolescents relative to normal controls. However, compared with healthy adolescents, eight regions involved in visual, language, and auditory processing were identified as hubs only present in prelingually deaf adolescents. These findings revealed reorganization of brain functional networks occurred in prelingually deaf adolescents to adapt to deficient auditory input. © 2016 Wenjing Li et al.


Li W.,Beijing University of Technology | Li W.,CAS Institute of Automation | Li W.,Beijing Key Laboratory of Computational Intelligence and Intelligent System | Li J.,Capital Medical University | And 6 more authors.
Restorative Neurology and Neuroscience | Year: 2015

Purpose: Previous studies have shown brain reorganizations after early deprivation of auditory sensory. However, changes of grey matter connectivity have not been investigated in prelingually deaf adolescents yet. In the present study, we aimed to investigate changes of grey matter connectivity within and between auditory, language and visual systems in prelingually deaf adolescents. Methods:We recruited 16 prelingually deaf adolescents and 16 age-and gender-matched normal controls, and extracted the grey matter volume as the structural characteristic from 14 regions of interest involved in auditory, language or visual processing to investigate the changes of grey matter connectivity within and between auditory, language and visual systems. Sparse inverse covariance estimation (SICE) was utilized to construct grey matter connectivity between these brain regions. Results: The results show that prelingually deaf adolescents present weaker grey matter connectivity within auditory and visual systems, and connectivity between language and visual systems declined. Notably, significantly increased brain connectivity was found between auditory and visual systems in prelingually deaf adolescents. Conclusions: Our results indicate "cross-modal" plasticity after deprivation of the auditory input in prelingually deaf adolescents, especially between auditory and visual systems. Besides, auditory deprivation and visual deficits might affect the connectivity pattern within language and visual systems in prelingually deaf adolescents. © 2015-IOS Press and the authors. All rights reserved.


Yang G.,Beijing University of Technology | Yang G.,East China Jiaotong University | Qiao J.F.,Beijing University of Technology | Qiao J.F.,Beijing Key Laboratory of Computational Intelligence and Intelligent System
Applied Soft Computing Journal | Year: 2014

Spatial architecture neural network (SANN), which is inspired by the connecting mode of excitatory pyramidal neurons and inhibitory interneurons of neocortex, is a multilayer artificial neural network and has good learning accuracy and generalization ability when used in real applications. However, the backpropagation-based learning algorithm (named BP-SANN) may be time consumption and slow convergence. In this paper, a new fast and accurate two-phase sequential learning scheme for SANN is hereby introduced to guarantee the network performance. With this new learning approach (named SFSL-SANN), only the weights connecting to output neurons will be trained during the learning process. In the first phase, a least-squares method is applied to estimate the span-output-weight on the basis of the fixed randomly generated initialized weight values. The improved iterative learning algorithm is then used to learn the feedforward-output-weight in the second phase. Detailed effectiveness comparison of SFSL-SANN is done with BP-SANN and other popular neural network approaches on benchmark problems drawn from the classification, regression and time-series prediction applications. The results demonstrate that the SFSL-SANN is faster convergence and time-saving than BP-SANN, and produces better learning accuracy and generalization performance than other approaches. © 2014 Elsevier B.V. All rights reserved.


Zuo G.,Beijing University of Technology | Zuo G.,Beijing Key Laboratory of Computational Intelligence and Intelligent System | Liu Y.,Beijing University of Technology | Wang X.,Beijing University of Technology
Proceedings of the World Congress on Intelligent Control and Automation (WCICA) | Year: 2016

This paper presents a design of a bionic robot inspired by the hopping locomotion of kangaroo including the hopping and transmission mechanisms. The jumping movement can be realized by the hopping mechanism while the transmission mechanism is used to transmit energy from the power source to the spring and release energy when necessary. The kinematic equations of the robot at take-off phase are established. Characteristics of the displacement and the velocity of the center of mass and supporting force during take-off are analyzed as well as the energy conversion efficiency of the robot. The simulation results show that the mechanical structure of the kangaroo-bionic hopping robot is well designed to meet the bionic hopping characteristics and has a high energy conversion efficiency. © 2016 IEEE.


Zuo G.,Beijing University of Technology | Zuo G.,Beijing Key Laboratory of Computational Intelligence and Intelligent System | Wu C.,Beijing University of Technology
Proceedings of the 28th Chinese Control and Decision Conference, CCDC 2016 | Year: 2016

This paper proposes a heuristic Monte Carlo Tree Search (MCTS) method for a computer game, Surakarta chess. The proposed method uses heuristic knowledge to guide the search procedure heading to a better solution. In our implementation we use a CPU based root parallelization technology which permits many instances running simultaneously. The reliability of our method has been proved to be considerably effective by experimental results. © 2016 IEEE.

Loading Beijing Key Laboratory of Computational Intelligence and Intelligent System collaborators
Loading Beijing Key Laboratory of Computational Intelligence and Intelligent System collaborators