The increased interest in discovering and managing the underlying, valid, insightful, and valuable knowledge from diverse, massive, complex, and fast changing data has stimulated the study and applications of many mathematical theories (for instance, topology, computational geometry, differential equations, fluid dynamics, and quantum statistics). Various methods and techniques have been proposed to solve problems in designing, analyzing, and implementing data-intensive and knowledge-based information systems. This special session aims at gathering researchers, engineers, and others interested in exploring mathematical methods for scientifically analyzing data and systematically managing knowledge to discuss theoretical and practical challenges and to present recent research progress and results dedicated (not exhausted) in
Abstract: Recently, computer-aided assessment (CAA) systems have been used for mathematics education, with some CAA systems able to assess learner's answers using mathematical expressions. However, the standard input method for mathematics is cumbersome for novice learners. In 2011, Professor Fukui proposed a new mathematical input method similar to the ones used for inputting Japanese characters in many operating systems. This method allows users to input mathematical expressions using colloquial-style linear strings. Therefore, users don’t need to learn new command syntax. However, they must convert each element contained in the colloquial-style mathematical string in order, from left to right. To improve this situation, we have proposed a predictive algorithm that uses a structured perceptron mainly for natural language processing as a previous study. This algorithm achieved an accuracy of up to 95.0% in terms of the top ten ranking. However, we also found that the score parameter continues to rise linearly during the learning process.
In this study, we propose a predictive algorithm with a prediction accuracy of 96.2% for the top ten ranking by improving a previous algorithm in terms of a structured perceptron for stable score parameter learning. In our evaluation of the prediction accuracy with a training sets of size 700, there is no statistically significant difference between the results for the proposed algorithm and those for the previous algorithm. Additionally, the mean scores among the correct expressions in the test dataset for each training number with the algorithm only increases at a rate of log n.
Abstract: In automotive configuration the question of the minimum number of test vehicles which cover all relevant options is desirable to avoid unnecessary cost. This problem is related to the minimum set cover problem, but with the restriction that we cannot enumerate all vehicle variants since the number of vehicle variants for a model type is too large. In this work we describe different use cases in the context of automotive configuration and give formal problem descriptions. We develop approximated and exact algorithms.
Abstract: The increased dimensionality of genomic and proteomic data produced by microarray and mass spectrometry technology makes testing and training of general classification method difficult. Special data analysis is demanded in this case and one of the common ways to handle high dimensionality is identification of the most relevant features in the data. Wrapper feature selection is one of the most common and effective techniques for feature selection. Although efficient, wrapper methods have some limitations due to the fact that their result depends on the search strategy. In theory when a complex search is used, it may take much longer to choose the best subset of features and may be impractical in some cases. Hence we propose a new wrapper feature selection for big data based on a random search using genetic algorithm and prior information. The new approach was tested on 2 biological dataset and compared to two well known wrapper feature selection approaches and results illustrate that our approach gives the best performances.
Abstract: Biological data bases are characterized by a very large number of features and a few instances which make classication more difficult and time consuming. This problem can be solved using feature selection approach. The Filter feature selection method ranks features according to their significance level. Then it selects the most significant features and discards the rest. The discarded features may provide some useful information and could be useful to further consideration. Hence, we propose a new feature selection method that uses these eliminated features in order to increase the classication performance and avoid the curse of dimensionality. The new approach is based on the idea of transforming the value of the similar features into new instances for the retained features. We aim to reduce the feature space by performing features selection and increasing the learning space in creating new instances using the redundant features.