CLASSIFICATION OF EYE IMAGES IN HEALTHCARE SYSTEMS USING ACTIVE LEARNING

The objective of active learning is to pick the most insightful manual marking cases. Most previous academic work centered on choosing one unmarked example for every iteration. This may be ineffective because, with any example classified, the classification model must be retrained. In this article, we propose a batch mode active learning framework which applies the Fisher information matrix for the measurement of general information for a number of unsheet examples. For the proposed system, the biggest calculating problem is the effective identification of the subset of examples that have the most insightful information on the existing classification model in general. We propose an effective gullible algorithm based on characteristics of submodular functions in order to overcome the computational complexity.