Masafumi Matsuhara

A novel distributed genetic algorithm implementation with variable number of islands

Genetic algorithm (GA) has some inherent drawbacks which become apparent while trying to solve complex multimodal problems. They are slow and the efficiency depends on parameter values. Some methods were proposed for alleviating these problems. But they did not address all the drawbacks. In this work, we propose a new distributed implementation strategy named variable island GA (VIGA), where the number of islands vary. In VIGA, where the number of individuals in every island is 2, the parameter population size in an island is fixed. Other parameters like number of islands, crossover/mutation probabilities, also need not be set. As the generation progresses, islands are created or erased based on the convergence status of searching in each island. Experiments were done with different function optimization problems. For all experiments VIGA delivered better or at least as good results as obtained by other competitive algorithms, at the expense of less computation and communication costs.

Mobility pattern learning and route prediction based location management in PCS network

Mobile host (MH) has to be tracked in personal communication service (PCS) network, for which update and paging signals are required. The number of PCS network subscribers skyrocketed in recent years. To reuse channels over a distance, cell size is reduced and the number of cell crossing by user is becoming high. That makes optimal use of paging and update signal very important. In fact, most MH has unique movement profile, that contains the information of time, route, direction, etc., which is possible to learn and used to predict location. In this paper, we propose mobility pattern based location management scheme using the movement profile. Mobility pattern is learned and system will page only the restricted probable area. We compared the proposed scheme with distance-based location management. Improved cost saving is achieved.

A Japanese Input Method for Mobile Terminals Using Surface EMG Signals

The common use of mobile terminals is for text input. However, mobile terminals cannot be equipped with sufficient amount of keys because of the physical restrictions. To solve this problem we developed an input method using surface electromyogram (sEMG), treating arm muscle movements as input signals. This method involves no physical keys and can be used to input Japanese texts. In our experiments, the system was capable of inputting Japanese characters with a finger motion recognition rate of approximately 80%.

Distance based location management in cellular PCS network ? a critical study

A distance-based approach is an efficient strategy for location management of mobile hosts in a cellular PCN. But what is the optimum distance for distance-based location management? Does it depend on cell-size, and/or the movement behavior of an individual user, depending on her movement pattern, more efficient? Or should a simple fixed distance for all users be adopted? We have shown that a fixed distance of 4 is nearly optimum for all the mobile hosts, irrespective of a range of variation of their mobility behavior, and the cell size of the network. A more efficient adaptive distance algorithm, where distance is changed according to mobility behavior, is proposed.

Evaluation of Number-Kanji Translation Method of Non-Segmented Japanese Sentences Using Inductive Learning with Degenerated Input

Our proposed method enables us to promptly and easily input Japanese sentences into a small device. All the keys for input are only 12 keys, which axe 0, 1,..., 9, * and #. Therefore, we are able to input one Kana character per one keystroke. Furthermore, the system based on our method automatically generates the dictionary adapted to the target field because the system automatically acquires words by using inductive learning. The system is improved by its own learning ability.

Effectiveness of context-aware character input method for mobile phone based on artificial neural network

Opportunities and needs are increasing to input Japanese sentences on mobile phones since performance of mobile phones is improving. Applications like E-mail, Web search, and so on are widely used on mobile phones now. We need to input Japanese sentences using only 12 keys on mobile phones. We have proposed a method to input Japanese sentences on mobile phones quickly and easily. We call this method number-Kanji translation method. The number string inputted by a user is translated into Kanji-Kana mixed sentence in our proposed method. Number string to Kana string is a one-to-many mapping. Therefore, it is difficult to translate a number string into the correct sentence intended by the user. The proposed context-aware mapping method is able to disambiguate a number string by artificial neural network (ANN). The system is able to translate number segments into the intended words because the system becomes aware of the correspondence of number segments with Japanese words through learning by ANN. The system does not need a dictionary. We also show the effectiveness of our proposed method for practical use by the result of the evaluation experiment in Twitter data.

Effectiveness for Machine Translation Method Using Inductive Learning on Number Representation

On our proposed method, source language is translated into target language via Number Representation. A text in the source language is translated into a number representation text. The number representation text is the number string corresponding to the original source language text. The number representation text is translated into a number representation text for the target language. The number representation text is translated into a text in the target language. The text is the translation result finally. A number representation text is more abstract than the original text because the number representation text corresponds to several texts. The system based on our proposed method is able to acquire more translation rules on number representation than that on the original text by Inductive Learning. Moreover, the system disambiguates number representation by its own adaptability. In the experiment, the correct translation rate for our proposed method is higher than that for the method without number representation. Thus, it is proved that our proposed method is more effective for machine translation.

A novel ranking method of web search result using clustering and concordance count

In recent years, information on World Wide Web is exploding. Search Engine robots are used to search information on World Wide Web. However, a robot type search engine has a few problems. One problem is that, it is difficult for user to come up with an appropriate query for getting search results she/he intends. Moreover, it is difficult for users to understand contents of search results because a robot type search engine outputs too many search results in a long list format. To resolve these problems, many systems classify a robot type search engine results into clusters. Clusters are labeled and those labels are shown to the user. Cluster labels need to be appropriate words for the web site within the cluster. We have proposed a labeling method using concordance count. First, web search results are obtained by a query input, and the result is classified into clusters. We used our proposed method to assign proper labels to those clusters. To ensure that we use a novel method. We find the set of websites resulted from AND-query using an original query word and the cluster label. If this set and the member of the cluster are common, we say that the concordance count is high. If the concordance count is high, the cluster label is assigned high weight. Finally, we evaluate the accuracy of our proposed method by simulation experiments.

An Efficient Method to Set RBF Network Paramters Based on SOM Training

Radial Basis Function (RBF) Network is popularly used for solving pattern recognition problems. The training of RBF Network is faster compared to multi layer perceptron using error backpropagation. However, the RBF Network uses the pseudo inverse matrix to calculate weights from the hidden layer to the output layer. Thus calculation cost increases when the number of data and the number of hidden units increase. In addition, in RBF Network the decision of optimum number of hidden units is difficult. It is also more prone to overtraining, needing repeated train and test cycles to ascertain a proper number of the Network hidden units, so that generalization performance is good. In this work, we propose a technique to set up RBF network parameters which is fast, as well as the number of hidden units are automatically determined. We start with training a Self-Organizing Maps (SOM), which is a unsupervised training, though our samples are labeled. SOM can find the distribution of data in multidimensional space, and map it on a two dimensional display. The results of SOM network is used to calculate the RBF parameters. It is shown by experiments that using the proposed method, RBF network parameters can be determined much faster compared to existing technique. Moreover, the recognition rate for the test data was higher, showing better generalization performance.

Improvement of paging cost by updating using paging delay divergence

Distance-based approach is most efficient among the three strategies, namely time-based, zone-based, and distance based, used for location management of mobile hosts (MHs) in a cellular PCN. In this work, we made one important observation about the optimum distance and then proposed a new updating strategy. When the MHs are static, depending on where they had last updated, the paging cost and delay could be high for the incoming calls during the static period. To avoid this situation we use the variance of paging delay as a measure of the mobility of the host. We have shown that using this parameter as an indicator for updating, we can reduce the paging cost to a great extent. The improvement is shown by simulation.