Extended Expectation Maximization for Under-Fitted Models

Abstract

<p>In this paper, we generalize the well-known Expectation Maximization (EM) algorithm using the α− divergence for Gaussian Mixture Model (GMM). This approach is used in robust subspace detection when the number of parameters is kept small to avoid overfitting and large estimation variances. The level of robustness can be tuned by the parameterα. When α→1, our method is equivalent to the standard EM approach and for α <1 the method is robust against potential outliers. Simulation results show that the method outperforms the standard EM when it comes to mismatches between noise models and their realizations. In addition, we use the proposed method to detect active brain areas using collected functional Magnetic Resonance Imaging (fMRI) data during task-related experiments.</p>

Filtering Empty Video Frames for Efficient Real-Time Object Detection

“…In this paper, we focus on enhancing the efficiency and scalability of generic object detection. Small object detection [61,62] is important for remote (e.g., aerial/maritime) sensing and medical image analysis, but it has not been considered. Improving both the accuracy and efficiency of small object detection remains an open problem that requires more in-depth research.…”

“…A potential extension in the future would allow a user to reconfigure objects of interest according to the application at hand. For example, a user can specify birds, drones, and other features to identify drones in a restricted flight zone [61,63]. An end-to-end framework that enables the configuration, model design, training/fine-tuning, and efficient real-time inference will be desirable.…”