Papers - YAMAMORI Kunihito
-
Construction of GMDH-based Prediction model using GA
Yoshihara Ikuo, Ohta Kouji, Yamamori Kunihito
宮崎大學工學部紀要 32 315 - 320 2003.7
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：宮崎大学
Abstract ###Time series prediction by combining GMDH and Genetic Algorithm (GA) is developed. GMDH ###combines a simple nonlinear method and built model using GA automatically. The proposed model is ###compared with GP-based model with bench marks of Mackey-Glass chaos, and sunspots. The prediction ###methods are evaluated from the viewpoint of minimum prediction error and average prediction error. The ###proposed model is able to obtain the result better than other methods, or is the same than other methods.
-
A New Data Structure for Lin-Kernighan Traveling Salesman Heuristic
Nguyen Hung Dinh, Yoshihara Ikuo, Yamamori Kunihito, Yasunaga Moritoshi
宮崎大學工學部紀要 32 303 - 308 2003.7
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：宮崎大学
Abstract ###The Lin-Kernighan (LK) heuristic is one of the most effective and efficient algorithms for the ###Traveling Salesman Problem (TSP). However, the LK heuristic is quite complicated and has many ###choices for implementing it. Especially, the data structure for tour representation plays an important ###role in the LK's performance. Traditionally, binary trees (including splay trees) are asymptotically ###the best tour representation. Empirically, however, they are utilized only for solving problems with ###more than one million cities due to the large overhead. Arrays are suitable for solving problems ###having up to a thousand cities and two-level trees are used for the problems with a thousand to a ###million cities. This paper proposes three-level trees as a new data structure. Although this structure is ###asymptotically not better than the existing ones, it perform empirically better than the existing ones ###in the range being investigated in this study (from 10(3乗) to 10(6.5乗) cities).
-
Minimizing the Number of Cameras to Take Pictures of All the Indoor Landscapes Based on GA
Yoshihara Ikuo, Arimura Kazuhiko, Yamamori Kunihito
宮崎大學工學部紀要 32 309 - 314 2003.7
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：宮崎大学
Abstract ###Minimizing the number of cameras is search for optimal all location for taking indoor ###pictures all the landscapes. It is necessary to take the conditions of a camera and walls into ###consideration. We proposed a GA-based method to minimize the number of cameras to take ###pictures all the indoor landscapes. We used an actual floor map, and experimented based on ###Genetic Algorithm. It was able to ask for the number of a camera considered to be close to the ###minimum as a result.
-
Mosaic Face Image Recognition on Multi-Layer Neural Network
Yamamori Kunihito, Nogawa Reo, Yoshihara Ikuo
宮崎大學工學部紀要 32 351 - 355 2003.7
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：宮崎大学
Face image recognition is an impotant technology for security,communication area,etc.. In this reserch,###we try to show the optimal parameters in multi-layer neural network for mosaic face image recognition.###By using of mosaic face images,the amount of image dara can be reduced,and it can also avoid###the affect of noise.Through our experiments,a multi-layer neural network showed 98.7% of recognition###on 8 x 8 mosaic images.
-
Performance of Parallel Back-Propagation Algorithm on PC Cluster System
Yamamori Kunihito, Tosa Yasushi, Yoshihara Ikuo
宮崎大學工學部紀要 32 345 - 349 2003.7
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：宮崎大学
ABSTRACT ###Neural network has been applied for many fields. However, real-world problems need large scale ###neural network, and enormous calculation time is also required for the training of neural network. So we ###made sample large-scale problem and parallelize it on used the computer cluster. This paper deals with ###parallel neural network on PC cluster system. We valuated the performance of the training-set parallel ###model on the PC cluster system. On the classification problem, the training-set parallel model was 6 ###cluster elements achieve 1.79 times faster than serial model.
-
Zhu Hanxi, Yoshihara Ikuo, Yamamori Kunihito, Yasunaga Moritoshi
宮崎大學工學部紀要 32 295 - 302 2003.7
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：宮崎大学
Prediction of protein secondary structure is considered as an important step towards elucidating###its three-dimensional structure, as well as its function. We have developed a multi-modal neural###network for predicting protein secondary structure. The prediction is based on the frequency profile of###multiple sequences alignment and the position specific scoring matrices (PSSM) generated by###BLOCK. The multi-modal neural network is composed of two steps: The first step is to develop three###neural networks to predict the secondary structure states of proteins: α-helix, β-sheet and###non-regular structure respectively. The single-state prediction neural networks use a local input###window of consecutive amino acids to predict the secondary structure state of the amino acid located###at the center of the input window; The second step is to develop a decision neural network to combine###all of the single-state predictions to obtain an overall prediction on three states. This method gives an###overall accuracy of 67.8% when using seven-fold cross-validation on a database of 126###non-homologous proteins. To improve the accuracy further, majority decision is introduced to each###network for single-state prediction in the first step. By using majority decision, the overall accuracy is###improved to 70.2% with corresponding Matthews' correlation coefficients Cα =0.61, Cβ=0.48.
-
Parallel Genetic Programming on PC Cluster System Reviewed
K.Yamamori, S.Matsumoto, T.Ohta, I.Yoshihara
The Seventh World Multiconference on Systems, Cybernetics and Informatics 7 259 - 264 2003.7
Language：English Publishing type：Research paper (international conference proceedings)
-
Quantitative Comparison of Defect Compensation Schemes for Multi-Layer Neural Networks with Flip-Link Defects Reviewed
K.Yamamori, K.Takahashi, I.Yoshihara
Proc. The Eighth International Symposium on Artificial Life and Robotics 2 484 - 487 2003.1
Language：English Publishing type：Research paper (international conference proceedings)
-
A Multi-modal Neural Net-work with Single-state Predictions for Protein Sec-ondary Structure Reviewed
H.Zhu, I.Yoshihara, K.Yamamori,M.Yasunaga
Proc. The Eighth International Symposium on Artificial Life and Robotics 2 475 - 478 2003.1
Language：English Publishing type：Research paper (international conference proceedings)
-
Improved GA-based method for multiple protein sequence alignment
Nguyen H., Yamamori K., Yoshihara I., Yasunaga M.
2003 Congress on Evolutionary Computation, CEC 2003 - Proceedings 3 1826 - 1832 2003
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：2003 Congress on Evolutionary Computation, CEC 2003 - Proceedings
In previous work, we have proposed a parallel hybrid genetic algorithm (PHGA) which can find high quality solution from the mathematical viewpoint for the multiple protein sequence alignment. We present new improvements to the PHGA. Local alignment information is added to the weighted sum-of-pairs objective function to achieve better alignment from the biological viewpoint. We also extend our method to run in parallel on a cluster of machines instead of a multi-processor machine to speed it up. © 2003 IEEE.
-
Greedy Genetic Algorithms for Symmetric and Asymmetric TSPs Reviewed
D.H.Nguyen, I.YoshiharaK.Yamamori, M.Yasunaga
IPSJ Trans. on Mathematical Modeling and its Applications 43 ( SIG10 ) 165 - 175 2002.12
Language：English Publishing type：Research paper (scientific journal)
-
Aligning Multiple Protein Sequences by Parallel Hybrid Genetic Algorithm Reviewed
D.H.Nguyen, I.Yoshihara,K.Yamamori, M.Yasunaga
Genome Informatics 2002 ( 13 ) 123 - 132 2002.12
Language：English Publishing type：Research paper (scientific journal)
-
Comparison with Defect Compensation Methods for Feed-forward Neural Networks Reviewed
K.Takahashi, K.Yamamori, S.Horiguchi, I.Yoshihara
Proc. 2002 Pacific Rim International Symposium on Dependable Computing 293 - 300 2002.12
Language：English Publishing type：Research paper (international conference proceedings)
-
Prediction of Protein Secondary Structure Based on Multi-Modal Neural Networks
H.Zhu, I.Yasunaga, K.Yamamori, M.Yasunaga
Proc. of the 4th Asia-Pacific Conference on Simulated Evolution and Learning crl-2132 2002.12
Language：English Publishing type：Research paper (international conference proceedings)
-
Greedy Genetic Algorithms for Symmetric and Asymmetric TSPs (特集 進化的計算)
HUNGDINHNGUYEN, IKUO YOSHIHARA, KUNIHITO YAMAMORI, MORITOSHI YASUNAGA
情報処理学会論文誌数理モデル化と応用（TOM） 43 ( 10 ) 165 - 175 2002.11
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：一般社団法人情報処理学会
This paper presents new enhancements to a multi-population GENITOR-type Genetic Al-gorithm (GA)for solving symmetric and asymmetric Traveling Salesman Problems (TSPs). First improvements to the greedy subtour crossover are proposed so that it works more effectively at the stage of highly t individuals.Next local search heuristics are combined with GA to compensate for its lack of local search ability.The powerful Lin-Kernighan heuristic is used for symmetric TSPs and the fast 3-Opt heuristic is used for asymmetric TSPs. Various symmetric and asymmetric TSP benchmarks taken from the TSPLIB are used to validate the method.Experimental results show that the proposed method can nd optimal solutions for problems ranging in size up to 3795 cities in a reasonable computing time. From the viewpoint of quality of solution these results are the best so far obtained by applyingGA to the TSP.This paper presents new enhancements to a multi-population GENITOR-type Genetic Al-gorithm (GA)for solving symmetric and asymmetric Traveling Salesman Problems (TSPs). First,improvements to the greedy subtour crossover are proposed so that it works more effectively at the stage of highly t individuals.Next,local search heuristics are combined with GA to compensate for its lack of local search ability.The powerful Lin-Kernighan heuristic is used for symmetric TSPs and the fast 3-Opt heuristic is used for asymmetric TSPs. Various symmetric and asymmetric TSP benchmarks taken from the TSPLIB are used to validate the method.Experimental results show that the proposed method can nd optimal solutions for problems ranging in size up to 3795 cities in a reasonable computing time. From the viewpoint of quality of solution,these results are the best so far obtained by applyingGA to the TSP.
-
Methods to Extract Exon Regions from DNA sequence : Development of Multi-modal Neural Network
KAMIMAI Yoshiyuki, YOSHIHARA Ikuo, YAMAMORI Kunihito, YASUNAGA Moritoshi
FAN' 02 12 385 - 390 2002.11
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：The Japan Society of Mechanical Engineers
Human genome consists of 3 billion base pairs, which are divided into several kinds of regions for example promoter, exon, and intron. Exons are the protein-coding DNA sequence, but introns are not. It is important to extract exons efficiently. This paper presents a multi-modal neural network to identify exon-intron boundaries and intron-exon boundaries in DNA sequences. The multi-modal neural network is composed several multi-layer neural networks and a judgment module, and gives the best identification rate (Sensitivity : Sn) 97.1%.
-
GP Based Method for Identifying Exon Region in DNA Sequences Reviewed
T.Ohta, I.Yoshihara, K.Yamamori,M.Yasunaga
Proc. The 4th Asia- Pacific Conference on Simulates Evolution and Learning crl1333 2002.11
Language：English Publishing type：Research paper (international conference proceedings)
-
Genetic Algorithm Approach to University Timetabling Reviewed
N.Koizumi, K.Yamamori, I.Yoshihara
Proc. The 4th Asia- Pacific Conference on Simulates Evolution and Learning, crl341(CD-ROM) crl-1341 2002.11
Language：English Publishing type：Research paper (international conference proceedings)
-
Compound Defect Compensation on Multi-Layer Neural Networks by Partial Retraining Scheme Reviewed
K.Yamamori, T.Abe, S.Horiguchi,I.Yoshihara
Proc. The Third International Conference on Parallel and Distributed Computing, Applications and Technologies 403 - 408 2002.9
Language：English Publishing type：Research paper (international conference proceedings)
-
A GA-based method for multiple protein sequence alignment
Nguyen Hung Dinh, Yoshihara Ikuo, Yamamori Kunihito, Yasunaga Moritoshi
宮崎大學工學部紀要 31 347 - 354 2002.7
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：宮崎大学
This paper presents a parallel hybrid genetic algorithm (GA) for solving the sum-of-pairs multiple protein sequence alignment. A new chromosome representation and its corresponding genetic operators are proposed. A multi-population GENITOR-type GA is combined with local search heuristics. It is then extended to run in parallel on a multiprocessor system for speeding up. Experimental results of benchmarks from the BAliBASE show that the proposed method is superior to MSA, OMA, and SAGA methods with regard to quality of solution and running time. It can be used for finding multiple sequence alignment as well as testing cost functions.