Learning Dropout Parameters for Convolutional Neural Networks

Abstract

Deep learning-based approaches have been paramount in the last years, mainly due to their outstanding results in several application domains, that range from face and object recognition to handwritten digits identification. Convolutional Neural Networks (CNN) have attracted a considerable attention since they model the intrinsic and complex brain working mechanism. However, one main shortcoming of such models concerns their overfitting problem, which prevents the network from predicting unseen data effectively. In this proposal, we address this problem by means of proper selecting regularization parameters by means of meta-heuristic-driven techniques, which provide a simple and elegant solution to a number of optimization problems. The proposed approach will be validated in the context of nuclei detection in general-purpose cell images under the supervision of Prof. Gustavo Rohde, University of Virginia.