Data assimilation for the next generation: Automatic modeling and information detection
Finished by March 31, 2013
The Institute of Statistical Mathematics, Research Organization of Information and Systems, Department of Statistical Modeling, Associate professor
To acquire knowledge from large data, it is crucial to construct an appropriate model and understand a flow of information. It is, however, easier said than done. This study systematically develops methods for (1) making the model more efficient and less computationally expensive and (2) detecting a flow of information that is grasped by the model. These methods are based on the technique so-called "data assimilation," which synthesizes a model and data. The methods enable us to remodel the original model without the model developer and to conduct a dynamic analysis of the model output that may even exceed the amount of input data.