The increasing pervasiveness of intelligent mobile applications requires to exploit the full range of resources offered by the mobile-edge-cloud network for the execution of inference tasks. However, ...due to the heterogeneity of such multi-tiered networks, it is essential to make the applications' demand amenable to the available resources while minimizing energy consumption. Modern dynamic deep neural networks (DNN) achieve this goal by designing multi-branched architectures where early exits enable sample-based adaptation of the model depth. In this paper, we tackle the problem of allocating sections of DNNs with early exits to the nodes of the mobile-edge-cloud system. By envisioning a 3-stage graph-modeling approach, we represent the possible options for splitting the DNN and deploying the DNN blocks on the multi-tiered network, embedding both the system constraints and the application requirements in a convenient and efficient way. Our framework - named Feasible Inference Graph (FIN) - can identify the solution that minimizes the overall inference energy consumption while enabling distributed inference over the multi-tiered network with the target quality and latency. Our results, obtained for DNNs with different levels of complexity, show that FIN matches the optimum and yields over 65% energy savings relative to a state-of-the-art technique for cost minimization.
The challenge of reliable cyberphysical system (CPS) engineering requires design automation methods that fully exploit algorithmic and datadriven approaches to NP-hard problem solving. Full-stack ...synthesis tools that integrate design, implementation, and certification are crucial for developing future generations of CPSs with rigorous functionality and reliability guarantees.
The location planning of relief distribution centres (DCs) is crucial in humanitarian logistics as it directly influences the disaster response and service to the affected victims. In light of the ...critical role of facility location in humanitarian logistics planning, the study proposes a two-stage relief distribution location problem. The first stage of the model determines the minimum number of relief DCs, and the second stage find the optimal location of these DCs to minimize the total cost. To address a more realistic situation, restrictions are imposed on the coverage area and capacity of each DCs. In addition, for optimally solving this complex NP-hard problem, a novel two-phase algorithm with exploration and exploitation phase is developed in the paper. The first phase of the algorithm i.e., exploration phase identifies a near-optimal solution while the second phase i.e. exploitation phase enhances the solution quality through a close circular proximity investigation. Furthermore, the comparative analysis of the proposed algorithm with other well-known algorithms such as genetic algorithm, pattern search, fmincon, multistart and hybrid heuristics is also reported and computationally tested from small to large data sets. The results reveal that the proposed two-phase algorithm is more efficient and effective when compared to the conventional metaheuristic methods.
We study the problem of minimizing energy costs for content delivery in wireless heterogeneous networks by jointly optimizing caching and power control strategies. This can be equivalently cast as a ...problem of maximizing the joint caching and power gain subject to meeting minimum signal-to-interference-plus-noise ratio constraints. The offline version of this problem is NP-hard, but we show that there exist polynomial-time approximation algorithms producing solutions within a constant factor 1 - 1/ e from the optimal. We further provide an adaptive algorithm based on projected subgradient ascent over a concave relaxation of the expected joint caching and power gain, which yields the same approximation guarantee. We show that our proposed algorithm outperforms the alternating optimization method and other baseline algorithms in a number of network scenarios, in total power consumption and run time.
A profiling-based algorithm for exams’ scheduling problem Alrawashdeh, Tawfiq; Al-Moghrabi, Khaldun G.; Al-Ghonmein, Ali M.
International journal of electrical and computer engineering (Malacca, Malacca),
10/2023, Volume:
13, Issue:
5
Journal Article
Typically, the problem of scheduling exams for universities aims to determine a schedule that satisfies logistics constraints, including the number of available exam rooms and the exam delivery mode ...(online or paper-based). The objective of this problem varies according to the university’s requirements. For example, some universities may seek to minimize operational costs, while others may work to minimize the schedule's length. Consequently, the objective imposed by the university affects the complexity of the problem. In this study, we present a grouping-based approach designed to address the problem of scheduling the exam timetable. The approach begins by profiling the courses’ exams based on their requirements, grouping exams with similar requirements to be scheduled at the same time. Then, an insertion strategy is used to obtain the exam schedule while satisfying the imposed constraints of the targeted university. We applied this approach to the problem of exam scheduling at Al-Hussein Bin Talal University in Jordan and achieved a balanced exam schedule that met all the imposed constraints.
This paper presents a quality of experience (QoE)-guided content delivery framework in cache-enabled multi-unmanned aerial vehicle (UAV) networks, measured from the content delay index (CDI), which ...is formulated based on the latency involved in the delivery of contents requested by ground-based mobile users. The mobile users use the cellular devices to raise the content requests. Additionally, multiple UAVs act as aerial base stations to deliver the requested contents to ground-based mobile users. The proposed work aims to maximize the QoE. Additionally, the existing works do not consider the necessity of creating groups of ground-based mobile users based on geographic locations in order to provide contents seamlessly. Therefore, we construct an optimization problem to maximize the average CDI, which is proved as an NP-hard problem. Further, multiple UAVs incur the challenges of placing the UAVs in 3D space over the distributed ground-based mobile users. To solve this NP-hard problem, we propose QoE analysis in cache-enabled multi-UAV networks (QMUN). It functions in three steps. First, it creates clusters for the ground-based mobile users using the K-means algorithm and finds the optimized 2D positions of all the UAVs. Second, it finds the optimized heights of all UAVs before assigning them to the individual cluster. Finally, it infuses proactive content caching and optimizes the cache policy. Therefore, the delay associated with the content delivery is minimized, which leads to the generation of high value of CDI. The results of simulation depict the usefulness of QMUN in terms of average CDI, average throughput, average end-to-end delay, and user satisfaction. Specifically, the results depict that the average CDI of each user improves by <inline-formula><tex-math notation="LaTeX">11.04</tex-math></inline-formula>%, <inline-formula><tex-math notation="LaTeX">4.46</tex-math></inline-formula>%, and <inline-formula><tex-math notation="LaTeX">17.88</tex-math></inline-formula>% compared to the QMUN Without Cache (QMUN-WC), Greedy, and Random Selection (RS).
Abstract
The goal of this paper is to find a better linesearch technique and different combinations of optimization and linesearch algorithms. Line searchtechnique on the off chance that it looks for ...the base of a descent direction vector that, when processed iteratively with a sensible advance size.
Motivated by circumstances with severe computational resource limits (e.g., settings with strong constraints on memory or communication), this paper addresses the problem of concisely representing ...and processing information for estimation and planning tasks. In this paper, conciseness is a measure of explicit representational complexity: for filtering, we are concerned with maintaining as little state as possible to perform a given task; for the planning case, we wish to generate the plan graph (or policy graph) with the fewest vertices that is correct and also complete. We present hardness results showing that both filtering and planning are NP-hard to perform in an optimally concise way, and that the related decision problems are NP-complete. We also describe algorithms for filter reduction and concise planning, for which these hardness results justify the potentially suboptimal output. The filter-reduction algorithm accepts as input an arbitrary combinatorial filter, expressed as a transition graph, and outputs an equivalent filter that uses fewer I-states to complete the same filtering task. The planning algorithm, using the filter-reduction algorithm as a subroutine, generates concise plans for planning problems that may involve both nondeterminism and partial observability. Both algorithms are governed by parameters that encode tradeoffs between computational efficiency and solution quality. We describe implementation of both algorithms and present a series of experiments evaluating their effectiveness.
This paper studies a lane reservation problem for large sport events in big cities. Such events require organizers to deliver certain people and materials from athlete villages to geographically ...dispersed venues within a given travel duration. A lane reservation strategy is usually adopted in this circumstance to ensure that time-critical transportation tasks can be completed despite heavy urban traffic congestion. However, it causes negative impact on normal traffic. The problem aims to optimally select and reserve some lanes in a transportation network for the exclusive use of the tasks such that the total traffic impact is minimized. To solve the problem, we first develop an improved integer linear program. Then, its properties are analyzed and used to reduce the search space for its optimal solutions. Finally, we develop a fast and effective quantum-inspired evolutionary algorithm for large-size problems. Computational results on instances with up to 500 nodes in the network and 50 tasks show that the proposed algorithm is efficient in yielding high-quality solutions within a relatively short time.