Extracting informative representations from videos is fundamental for effectively learning various downstream tasks. We present a novel approach for unsupervised learning of meaningful representations from videos, leveraging the concept of image spatial entropy (ISE) that quantifies the per-pixel information in an image. We argue that local entropy of pixel neighborhoods and their temporal evolution create valuable intrinsic supervisory signals for learning prominent features. Building on this idea, we abstract visual features into a concise representation of keypoints that act as dynamic information transmitters, and design a deep learning model that learns, purely unsupervised, spatially and temporally consistent representations directly from video frames. Two original information-theoretic losses, computed from local entropy, guide our model to discover consistent keypoint representations; a loss that maximizes the spatial information covered by the keypoints and a loss that optimizes the keypoints’ information transportation over time. We compare our keypoint representation to strong baselines for various downstream tasks, e.g., learning object dynamics. Our empirical results show superior performance for our information-driven keypoints that resolve challenges like attendance to static and dynamic objects or objects abruptly entering and leaving the scene.

Large scale grids permit to share grid resources spread over different autonomous administrative sites in the internet. The rapid progress of grid systems opens the door for numerous companies to adopt this technology in their business development. This progress is characterized by the increasing openness and opportunity of resource-sharing across organizations in different domains. In the business context, these shared resources can be misused by some malicious users that can abuse the provided resources and make them behave maliciously to return wrong results and sabotage the jobs execution. The common technique used by most grid systems to deal with this problem is based on replication with voting. Nevertheless, these techniques rely on the assumption that the grid resources behave independently. They may be ineffective where a number of collusive resources collectively return the same wrong results of a job execution. In order to overcome this threat, we propose a reputation-based voting (RBV) approach, which investigates the trustworthiness of the grid resources through a reputation system, and then takes a decision about the results. In addition, the performance of our approach and other voting techniques, like m-first voting and credibility-based voting, are evaluated through simulation to perceive the effect of collusive grid resources on the correctness of the results. The obtained results show that our approach achieves a lower error-rate and better performance in terms of overhead.