e-Science and Computing

Abstract

The question addressed in this chapter is how to expand the scope for biometrics and how to add to existing knowledge and practice. Alternatively, the question addressed is how to combine the two kinds of minds Pascal talks about. The rapid increase expected in biometric data volumes is not matched by a commensurate increase in the quantity or quality of (data intensive) scientific research tools. To reach existing and future biometric challenges, we propose here the idea of agent-based middleware, driven by machine learning, to automate the process of data search, query formulation, workflow configuration and service composition, and collaborative reuse. This should permit biometrics to include a much larger volume and diversity of data in the discovery process than would otherwise be possible. The overall working theme here is moving from mere access to biometric data to information integration and scientific discovery, which should lead to more reliable face recognition systems. The end-user is faced with a seemingly inexhaustible search space of data collections and analysis tools. What this user needs is an intermediary (agent or middleware) architecture that recognizes the user's patterns of data usage and analysis, and then automatically assists the user with data discovery, navigation, retrieval, mining, fusion, and analysis. Precision and recall metrics validate the search and discovery capabilities of the proposed architecture.