A Study on Large Scale Data Analysis Method and Theoretical Evaluation based on Distance Metric Learning

The development in information technology highlighted the importance of knowledge discovery from enormous electronic data. We focus on the distance metric learning which is one of the methods of machine learning. In this study, we propose the regularization methods and the way to select the suitable training data in order to reduce computational complexity of distance metric learning. In addition, we propose the way to obtain multiple distance metrics and integrate those in order to improve the classification accuracy. Consequently, we clarify the effectiveness of our proposed methods. These methods can be used properly according to the characteristics of the analysis (e.g. to gain high performance, low computational complexity and so on). By using these method properly, it can be implemented various types of analysis.