Multimedia Devices Laboratory

/ Qiangfu Zhao / Professor
/ Jie Huang / Assistant Professor

The main stream in our lab is related to computational intelligence. The final 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 and evolutionary computation. Of course, results proposed in conventional symbol based artificial intelligence are also included.

1. Neuro-computation and evolutionary computation:

   The goal of this research is to develop a neural network model that is flexible enough to adapt changing environment, and also simple enough to be realized and used for practical applications. Of course, we cannot use a model if we do not study the learning algorithms corresponding to this model. Possible models include:

   • Modular neural networks
   • Neural network ensemble
   • Cellular automata and recurrent neural networks
   • Integration of symbolic and sub-symbolic models
   As for learning, we have proposed the following algorithms:
   • IEA: individual evolutionary algorithm
   • CoopCEA: cooperative co-evolutionary algorithms
   • EPNet: evolutionary programming neural net
   To verify and improve the models and learning algorithms proposed by us, we are also studying automatic design of decision rules and decision trees, pattern recognition, robot control, and data mining.

2. Supporting Researches:

   To apply our methods to pattern recognition and robot control, it is also necessary to get support from many other researches, such as signal and image processing, development of virtual environment, learning and evolution of agents living in it, and so on. Specifically, we are studying localization and recognition of sounds in natural environments, and trying to apply our result to automatic acquisition of robot control strategies.

Refereed Journal Papers

1. Huang, J. and Ohnishi, N. and Sugie, N., Echo Avoidance in a Computaional Model of the Precedence Effect. Speech Communication, vol.27, no.3-4, pp.223--233, 1999.

2. Huang, J. and Supaongprapa, T. and Terakura, I. and Wang, F. and Ohnishi, N. and Sugie, N., A Model Based Sound Localization System and Its Application to Robot Navigation. Robotics and Autonomous Systems, vol.27, no.4, pp.199-209, 1999.

1. Q. F. Zhao and M. Shirasaka., A study on evolutionary design of binary decision trees. Proc. IEEE Congress on Evolutionary Computation (CEC'99), pp.1988-1993, IEEE, July 1999.

   For pattern recognition, the decision trees (DTs) are more efficient than neural networks (NNs) for two reasons. First, the computations in making decisions are simpler. Second, important features can be selected automatically during the design process. However, the DTs are not adaptable. This problem can be avoided by mapping a DT to an NN. This mapping not only makes a DT adaptable, but also provides a systematic way for determining the NN structure. In addition, since the features are well selected, the NN obtained from this mapping may have much fewer connections than those designed directly. The key point here is to design a DT which is as small as possible. In this paper, we study the evolutionary design of the decision trees, and investigate some methods to improve the design efficiency.

2. Q. F. Zhao., Cooperative co-evolution of pattern recognition agents. Proc. 4-th Annual Meeting of Japan Association for Evolutionary Economics (JAFEE'2000), pp.166-169, JAFFE, Mar. 2000.

   A pattern recognition system can often be considered as a symbiotic system consisting of many expert agents. Although each agent may only be able to recognize a limited number of patterns, when they are put together, they may form a very powerful system. In this paper, we study the learning of such kind of pattern recognition agents based on the cooperative co-evolutionary algorithms (CoopCEAs). We will try to answer one question in this paper: how to evaluate the fitness of each agent ? Five evaluation methods are proposed and tested. Experimental results show that the CoopCEA outperforms the standard genetic algorithms only if the agents are evaluated properly.

3. Huang, J. and Ohnishi, N. and Guo, X. and Sugie, N., A Computational Model of Echo Avoidance in Human Audition. Proc. Int. Conf. Neural Networks and Brain, Beijing, CNNC, pp.61--66, Oct. 1999.

1. Tetsurou Komatsu., Improving GP to avoid the premature convergence. The Univ. of Aizu, 1999. Thesis Advisor: Q. F. Zhao.

2. Toru Tanigawa., Improving GP to generate smaller decision trees. The Univ. of Aizu, 1999. Thesis Advisor: Q. F. Zhao.

3. Tatsuya Namatame., Increasing generalization ability of GP through multiple environments. The Univ. of Aizu, 1999. Thesis Advisor: Q. F. Zhao.

4. Keiichi Hirano., Effect of evolution parameters in GP. The Univ. of Aizu, 1999. Thesis Advisor: Q. F. Zhao.

5. Takuya Kouyama. A comparison study of selection methods in GP. The Univ. of Aizu, 1999. Thesis Advisor: Q. F. Zhao.

6. Erina Itou., Feature selection based on GP. The Univ. of Aizu, 1999. Thesis Advisor: Q. F. Zhao.

7. Kume, K., Computational Implementation of Primary Cues for Auditory Stream Segregation. The Univ. of Aizu, 1999. Thesis Advisor: Jie Huang.

8. Yanagi, T., Implementation and Evaluation of a 3D Sound Model. The Univ. of Aizu, 1999. Thesis Advisor: Jie Huang.

9. Goto, T., A Weighted Cross-correlation Method for Sound Localization in Reverberant Environments. The Univ. of Aizu, 1999. Thesis Advisor: Jie Huang.

10. Utuno, Y., Computational Evaluation for the EA Model of the Precedence Effect. The Univ. of Aizu, 1999. Thesis Advisor: Jie Huang.