An Externally-Constrained Ising Clustering Method for Material Informatics

Abstract

Due to the recent advancement of data science, such as machine learning and big-data analysis, the approach using data science techniques has attracted attention even to develop new materials, called material informatics. In material informatics, clustering is one of the essential data processing techniques to understand thermophysical properties. Thus, clustering quality is a high priority to be considered. To improve clustering accuracy, this paper evaluates Ising-based clustering methods using an annealing machine. As an annealing machine minimizes the energy of an Ising model, the Ising-based clustering methods define the clustering as an Ising model to minimize the sum of intra-cluster distances among data. Since the non-Ising-based clustering methods conventionally used in materials informatics perform pseudo-optimization, the Ising-based clustering methods can achieve high clustering accuracy. The experimental results show that the Ising-based clustering method with externally-defined constraint achieves higher clustering accuracy with an affordable execution time than the conventional K-means clustering method.