Optimizing Fog Network Performance through Task Distribution in Deep Neural Networks

The transition from Internet of Things (IoT) devices to fog nodes in distributed computing addresses IoT's resource limitations, cost-effectiveness, and productivity issues by delegating duties. Despite constrained resources, fog nodes can address the increasing demands for mobile data and compute requirements in IoT applications and Deep Neural Networks (DNNs). Nonetheless, attaining optimal outcomes with conventional optimization techniques is difficult, notably when end-users are excluded from the decision-making process. Due to challenges in accessing original network parameters, deep learning methodologies encounter constraints in adjusting to novel situations. To tackle these difficulties, we present the Deep Neural Network-based Task Offloading Framework (DNNTOF), which utilizes deep learning's perceptual capabilities to execute adaptive, real-time offloading decisions. Simulations indicate that DNNTOF surpasses traditional deep learning methods, facilitating effective job offloading in dynamic and mobile IoT environments.