Qiangfu Zhao
Professor
Yong Liu
Associate Professor
|
Qiangfu Zhao Professor |
Yong Liu Associate Professor |
The main stream in our lab is related to computational intelligence. So far we have focused our study on three key words: recognition, learning and understanding. The goal of our research is to develop some learning models that are flexible enough to adapt changing environment, and also simple enough to be realized, interpreted and re-used. The ultimate goal is to design a system that can think, and decide what to do and how to grow-up based on its own thinking. For this purpose, many approaches have been studied - e.g., neuro-computation, evolutionary computation, reinforcement learning, awareness computing, and so on. Of course, results proposed in conventional symbol based artificial intelligence are also included.
In 2011, we successfully co-organized the IEEE International Conference on Awareness Science and Technology (iCAST2011) with Dalian University of Technology. iCAST2011 was technically co-sponsored by IEEE Computational Intelligence Society, IEEE Systems, Man, and Cybernetics Society, and Information Processing Society of Japan. We also prepared the IEEE International Conference on Awareness Science and Technology (iCAST2012). We have been trying to promote awareness technology through collaboration with Chaoyang University of Technology (CYUT) and Shandong Academy of Sciences (SDAS). Our dream is to propose a new and better approach to realization of artificial intelligence.
So far we have used or proposed the following learning models:
Neural network trees (NNTrees),
Nearest neighbor classification trees (NNC-Trees),
Support vector machines (SVMs),
Neural network ensembles,
Modular neural networks,
Cellular automata, and
Recurrent neural networks.
Based on the above learning models, we have proposed many new algorithms. Examples include:
IEA: individual evolutionary algorithm (also called the R4-rule),
CoopCEA: cooperative co-evolutionary algorithms,
EPNet: evolutionary programming neural net,
Evolutionary design of neural network trees,
Induction of NNC-Trees with the R4-rule,
Fast neural network for face detection,
Rough null space approach to face recognition,
Card holder authentication based on image fusion.
To verify and to improve the models and learning algorithms proposed so far, we have being studying on-line growing of neural network trees, evolution of neural network ensemble, evolutionary design of decision trees, and so on. Currently, we are very interested in applying our models and algorithms to solving practical problems related to producing a “safe, secure and healthy” society. Examples include: face detection, face/expression recognition, image compression and protection, text mining, IC card holder authentication, and so on.
Recently, we started a new research related to steganography. We have proposed a morphing based steganography technology that is useful for holder authentication and communication. We have filed several patents related to this research, and are trying to make some useful products in near future.
Q. F. Zhao and C.-H. Hsieh. Card User Awareness Based on Image Morphing. Journal of Computer Engineering and Science, 34(1):11-20, 1 2012.
Consciously or unconsciously, we are using many kinds of “cards ”in our daily lives. Among them, credit card, cash card, driving license, etc., are as important as our wallets. To prevent the cards from being illegally used, card user authentication is indispensable. A common practice for user authentication is to ask the user to provide some personal information. The user is usually considered legal if the personal information matches the one pre-stored in the card. However, if the personal information is leaked, some third party may use the card illegally. The purpose of this paper is to propose an image morphing based method for card user authentication. The fundamental idea is to hide the face image of the true card holder into some cover data through morphing, distribute the key information for restoring the face image into several different places, and restore the face image through de-morphing when the card is used. This method can provide good user awareness for authorized persons, and can protect the true card holder more effectively.
Jie Jie, Rung-Ching Chen, and Qiangfu Zhao. Document analysis system based on awareness learning. Journal of Advance Computational Intelligence and Intelligent Informatics, 15(9):1230-1240, 11 2011.
With the rapid growth of the Internet, it is natural that we want to handle and process the document not as pen-and-paper but as on-line information. To deal with such large amount of information efficiently, we need to classify the data into meaningful categories. Many machine learning based algorithms have been proposed for document classification. Based on these algorithms, various application systems have been developed. Such as spam mail filter, patent analyzer, and hot-topic retrieval system. Since different applications have different goals, we need different teacher signals even for the same data set. However, in real word, it is difficult to get enough teacher signals for every application. In this study, we describe a human behavior inspired awareness system for document analysis. This system can start learning with few even no teacher signals. It can learn and understand the user intention through interaction with the user. In this paper, we describe the structure of the proposed system, and show the basic steps for document analysis using the system.
Kuan-Chieh Hsiung Cheng-Hsiung Hsieh, Ching-Hua Liu and Qiangfu Zhao. Grey temporal error concealment. Grey Systems: Theory and Application, 1(2):138 147, 2011.
Purpose The purpose of this paper is to solve the pack loss problem of video transmitted over error-prone channels. Pack loss generally affects the visual quality of reconstructed frames significantly. Consequently, a grey approach to error concealment (EC) is proposed. Design/methodology/approach Note that missing information in error blocks can be found before and after the error frame. Thus, two adjacent error-free frames are utilized to conceal error blocks caused by packet loss. This paper presents an EC approach based on grey polynomial interpolation (GPI) which is called the GTEC. In the GTEC, the following stages are involved. First, error blocks due to packed loss are detected. Then, optimal reference blocks in adjacent frames are found through boundary matching algorithm (BMA). Finally, estimated blocks are obtained by the GPI. By replacing error blocks with the estimated blocks, EC is completed in the GTEC. Findings In the simulation, the proposed GTEC is compared with the EC scheme in H.264 and the BMA. With packet loss rates of 1, 3, 5, and 10 per cent, the proposed GTEC approach has better performance than EC schemes in H.264 and BMA, both in peak signalto-noise ratio and visual quality. Consequently, it provides an alternative where EC is required. Originality/value The value of GTEC proposed in this paper is not only in better performance but also in the originality to apply grey scheme, i.e. GPI, in the field of EC.
H. Wang, and S. Rahnamayan Z. Wu, Y. Liu, and M. Ventresca. Enhancing particle swarm optimization using generalized opposition-based learning. Information Sciences, 181(20):4699-4714, 2011.
Particle swarm optimization (PSO) has been shown to yield good performance for solving various optimization problems. However, it tends to suffer from premature convergence when solving complex problems. This paper presents an enhanced PSO algorithm called GOPSO, which employs generalized opposition-based learning (GOBL) and Cauchy mutation to overcome this problem. GOBL can provide a faster convergence, and the Cauchy mutation with a long tail helps trapped particles escape from local optima. The proposed approach uses a similar scheme as opposition-based differential evolution (ODE) with opposition-based population initialization and generation jumping using GOBL. Experiments are conducted on a comprehensive set of benchmark functions, including rotated multimodal problems and shifted large-scale problems. The results show that GOPSO obtains promising performance on a majority of the test problems.
Y. Liu. New discoveries in fast evolutionary programming. International Journal of Innovative Computing, Information & Control, 7(5(B)):2881 2896, 2011.
It had been stated in both the theoretical analysis and the empirical results on fast evolutionary programming (FEP) that long jumps of Cauchy mutations were the cause of the better performance of FEP on optimizing both unimodal and multimodal functions. Such a statement about long jumps of Cauchy mutations has been so widely held in the applications of evolutionary programming (EP) that the effectiveness of long jumps of Cauchy mutations has seldom been put in doubt. Through carefully examining the relationship between the step sizes of mutations and their performance, it has been discovered that not long jumps but short jumps with large variances among Cauchy mutations had led to the better performance of FEP than that of classical EP (CEP). Experimental results given in this paper show that effective Cauchy mutations in FEP had often had even shorter step sizes on average than effective Gaussian mutations in CEP, although the average step sizes of Cauchy mutations were much longer than those of Gaussian mutations. Effective mutations used here refer to those mutations that have generated better offspring than their parent solutions among all mutations at each generation. It has been further discovered that the same self-adaptation used in CEP and FEP had shown quite different behaviors on optimizing the same test functions from the same initial populations. These two discoveries shed light on why the shorter effective Cauchy mutations performed better than the longer effective Gaussian mutations, and how effective Cauchy mutations had had the shorter step sizes than effective Gaussian mutations.
Hayashi Hirotomo Jie Ji and Q. F. Zhao. Inducing Portable Neural Network Trees for Text Data Through DCAMC. In Proc. of the 4th International Conference on Human System Interaction,, pages 221-228, 2011.
An NNTree is a decision tree with each nonterminal node containing a neural network (NN). Our previous researches show that compared with neural networks, the NNtree can classify given data in a hierarchical structure which has very small system scale can can be applied to many PORTABLE DEVICE applications. However, for text data, the high dimensionality is a serious problem for induction of NNTrees since the system scale may still become too large and each NN spends too much time for training. To solve the problem, we have proposed discriminant multiple center (DMC) method. In this paper, we combined DMC method with comparative advantage (CA) based algorithm together and proposed discriminant comparative advantage based multiple center (DCAMC) method for inducing NNTrees. DCAMC is a two-stage approach, in which all data are first mapped to a lower dimensional space based on the comparative advantage law, and the LDA is then conducted on the mapped space. Experimental results on three popular databases show that DCAMC can produce NNTrees more efficiently than DMC method.
Q. F. Zhao and C.-H. Hsieh. Card User awareness based on linear subspace representation. In IEEE, editor, Proc. of IEEE International Conference on Systems, Man and Cybernetics, pages 227-232. IEEE, 10 2011.
Nowadays, “ cards ”are becoming part of our lives. To prevent the cards from being illegally used, card user authentication is indispensable. A common practice for user authentication is to use some secret information known or possessed by the true card holder. However, if the secret information is leaked, some malicious third party may still be able to use the card illegally. To solve the problem, this paper proposes a new authentication method based on linear sub-space representation. The basic idea is to hide the face image of the card holder into some other images, distribute the key information for restoring the face image into several places, and restore the face image when necessary. Compared with the methods we proposed earlier, the new method is more efficient, and is easier to implement.
Y. Liu. Balancing ensemble learning through error shift. In Proceedings of the Fourth International Workshop on Advanced Computational Intelligence, pages 351-358, Wuhan, Oct. 2011. IEEE Press.
In neural network learning, it has been often observed that some data have been learned extremely well while others have been barely learned. Such unbalanced learning often lead to the learned neural networks or neural network ensembles that could be too strongly biased on those learned-well data. The stronger bias could contribute to the larger variance and the poorer generalization on the unseen data. It is necessary to prevent a learned model from being strong biased especially when the model have unnecessary large complexity for the application. This paper shows how balanced ensemble learning could guide learning to being less biased through error shift, and create weak learners in an ensemble.
Y. Liu. Create weak learners with small neural networks by balanced ensemble learning. In Proceedings of the 2011 IEEE International Conference on Signal Processing, Communications and Computing, pages 733-736, Xi'an, Sept. 2011. IEEE Press.
It has been shown that as the number of weak learners in a majority voting model is increased so does its generalization if those weak learners are uncorrelated or negatively correlated. Although some learning algorithms including bagging and boosting have been developed to create such weak learners, learners trained by these learning algorithms are actually not so weak in many applications. This paper presents a simple balanced ensemble learning method for producing weak learners. The idea of balanced ensemble learning is to change the learning force in the training process so that the training data points near to the decision boundaries would push the decision boundaries further while the training data points far away from the decision boundaries would drag the decision boundaries to themselves. The experimental results suggest that balanced ensemble learning is able to create learners being both weak and negatively correlated.
Y. Liu. Impact of alternating current on the switch-on duration of optical switches by thin films with defect layers. In Proceedings of the 2011 IEEE International Conference on Signal Processing, Communications and Computing, pages 825-828, Xi'an, Sept. 2011. IEEE Press.
This paper analyze the impact of alternating current on the switch-on duration (SOD) of optical switches by thin films with defect layers. The optical switch is implemented with the structure (AB)5Ck(BA)5, in which the dielectric materials are
Y. Liu. Target shift awareness in balanced ensemble learning. In Proceedings of the 3rd International Conference on Awareness Science and Technology, pages 166-170, Dalian, Sept. 2011. IEEE Press.
In the balanced ensemble learning for a two-class classification problem, the target values are shifted between [1 : 0:5) or (0:5 : 0] instead of 1 and 0 in the learned error function. Such shifted error function could let the ensemble avoid from unnecessary further learning on the well-learned data points. Therefore, the learning direction could be shifted away from the well-learned data points, and turned to the other not-yet-learned data points. By shifting away from well-learned data and focusing on not-yet-learned data, a good balanced learning could be achieved in the ensemble. Through examining both individual learners and the combined ensembles, this paper is to explore how the target shift awareness could help to decide a decision boundary that is neither too close nor too further to all training samples.
X. Guo, Y. Toyoda, H. Li, J. Huang, S. Ding, and Y. Liu. Environmental sound recognition using time-frequency intersection patterns. In Proceedings of the 3rd International Conference on Awareness Science and Technology, pages 244-247, Dalian, Sept. 2011. IEEE Press.
Environmental sound recognition is an important function of robots and intelligent computer systems. In this research, we use a multistage perceptron neural network system for environmental sound recognition. The input data is a combination of timevariance pattern of instantaneous powers and frequency-variance pattern with instantaneous spectrum at the power peak, referred to as a time-frequency intersection pattern. Spectra of many environmental sounds change more slowly than those of speech or voice, so the intersectional time-frequency pattern will preserve the major features of environmental sounds but with drastically reduced data requirements. Two experiments were conducted using an original database and an open database created by the RWCP project. The recognition rate for 20 kinds of environmental sounds was 92%. The recognition rate of the new method was about 12% higher than methods using only an instantaneous spectrum. The results are also comparable with HMM-based methods, although those methods need to treat the time variance of an input vector series with more complicated computations.
Y. Liu. The Grant-In-Aid for Scientific Research Fund (Kakenhi), 2011—2013.
Mayuko Akatsuka. evolution of natural face images based on morphing. Master thesis, Graduate School of Computer Science and Engineering, March 2012.
Thesis Adviser: Q.F. Zhao
Kazutoshi Igarashi. A study on fast image morphing. Graduation thesis, School of Computer Science and Engineering, March 2012.
Thesis Adviser: Q.F. Zhao
Yutaro Minakawa. A Study on document based information retrieval. Graduation thesis, School of Computer Science and Engineering, March 2012.
Thesis Adviser: Q.F. Zhao
Yuta Hiraki. A study on user intention modeling. Master thesis, Graduate School of Computer Science and Engineering, March 2012.
Thesis Adviser: Q.F. Zhao
Tomoyuki Soeta. A study on image based information retrieval. Graduation thesis, School of Computer Science and Engineering, March 2012.
Thesis Adviser: Q.F. Zhao
Yuya Kaneda. A study on designing compact SMVs based on dimensionality reduction. Graduation thesis, School of Computer Science and Engineering, March 2012.
Thesis Adviser: Q.F. Zhao
Yuhi Miura. Increasing the quality of image morphing based on normalization. Graduation thesis, School of Computer Science and Engineering, March 2012.
Thesis Adviser: Q.F. Zhao
Kazutoshi Maeda. User authentication based on linear image approximation. Master thesis, Graduate School of Computer Science and Engineering, March 2012.
Thesis Adviser: Q.F. Zhao
Jie Ji. Fast document analysis algorithms based on the comparative advantage theory. Doctor thesis, Graduate School of Computer Science and Engineering, September 2011.
Thesis Adviser: Q.F. Zhao
Tomohiro Yoshida. Generation of Natural Images by Genetic Algorithms. Graduation thesis, School of Computer Science and Engineering, 2012.
Thesis Adviser: Y. Liu
Keisuke Watanabe. Comparisons between SVM and AdaBoost on Face Dectection. Graduation thesis, School of Computer Science and Engineering, 2012.
Thesis Adviser: Y. Liu
Takumi Emori. Reinforcement Learning for Maze Problems with Random Start. Graduation thesis, School of Computer Science and Engineering, 2012.
Thesis Adviser: Y. Liu
2012 THE UNIVERSITY OF AIZU ALL RIGHTS RESERVED