skip to main content

Search for: All records

Award ID contains: 1724227

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. Free, publicly-accessible full text available March 1, 2023
  2. Free, publicly-accessible full text available December 1, 2022
  3. Deep neural networks (DNNs) have became one of the most high performing tools in a broad rangeof machine learning areas. However, the multilayer non-linearity of the network architectures preventus from gaining a better understanding of the models’ predictions. Gradient based attributionmethods (e.g., Integrated Gradient (IG)) that decipher input features’ contribution to the predictiontask have been shown to be highly effective yet requiring a reference input as the anchor for explainingmodel’s output. The performance of DNN model interpretation can be quite inconsistent withregard to the choice of references. Here we propose an Adversarial Gradient Integration (AGI) methodthat integrates the gradients frommore »adversarial examples to the target example along the curve of steepestascent to calculate the resulting contributions from all input features. Our method doesn’t rely onthe choice of references, hence can avoid the ambiguity and inconsistency sourced from the referenceselection. We demonstrate the performance of our AGI method and compare with competing methodsin explaining image classification results. Code is available from https://github.com/pd90506/AGI.

    « less
  4. Characterizing computational demand of Cyber-Physical Systems (CPS) is critical for guaranteeing that multiple hard real-time tasks may be scheduled on shared resources without missing deadlines. In a CPS involving repetition such as industrial automation systems found in chemical process control or robotic manufacturing, sensors and actuators used as part of the industrial process may be conditionally enabled (and disabled) as a sequence of repeated steps is executed. In robotic manufacturing, for example, these steps may be the movement of a robotic arm through some trajectories followed by activation of end-effector sensors and actuators at the end of each completed motion.more »The conditional enabling of sensors and actuators produces a sequence of Monotonically Ascending Execution times (MAE) with lower WCET when the sensors are disabled and higher WCET when enabled. Since these systems may have several predefined steps to follow before repeating the entire sequence each unique step may result in several consecutive sequences of MAE. The repetition of these unique sequences of MAE result in a repeating WCET sequence. In the absence of an efficient demand characterization technique for repeating WCET sequences composed of subsequences with monotonically increasing execution time, this work proposes a new task model to describe the behavior of real-world systems which generate large repeating WCET sequences with subsequences of monotonically increasing execution times. In comparison to the most applicable current model, the Generalized Multiframe model (GMF), an empirically and theoretically faster method for characterizing the demand is provided. The demand characterization algorithm is evaluated through a case study of a robotic arm and simulation of 10,000 randomly generated tasks where, on average, the proposed approach is 231 and 179 times faster than the state-of-the-art in the case study and simulation respectively.« less
  5. In Vehicular Edge Computing (VEC) systems, the computing resources of connected Electric Vehicles (EV) are used to fulfill the low-latency computation requirements of vehicles. However, local execution of heavy workloads may drain a considerable amount of energy in EVs. One promising way to improve the energy efficiency is to share and coordinate computing resources among connected EVs. However, the uncertainties in the future location of vehicles make it hard to decide which vehicles participate in resource sharing and how long they share their resources so that all participants benefit from resource sharing. In this paper, we propose VECMAN, a frameworkmore »for energy-aware resource management in VEC systems composed of two algorithms: (i) a resource selector algorithm that determines the participating vehicles and the duration of resource sharing period; and (ii) an energy manager algorithm that manages computing resources of the participating vehicles with the aim of minimizing the computational energy consumption. We evaluate the proposed algorithms and show that they considerably reduce the vehicles computational energy consumption compared to the state-of-the-art baselines. Specifically, our algorithms achieve between 7% and 18% energy savings compared to a baseline that executes workload locally and an average of 13% energy savings compared to a baseline that offloads vehicles workloads to RSUs.« less
  6. Convolutional neural networks (CNNs) have achieved state-of- the-art performance on various tasks in computer vision. However, recent studies demonstrate that these models are vulnerable to carefully crafted adversarial samples and suffer from a significant performance drop when predicting them. Many methods have been proposed to improve adversarial robustness (e.g., adversarial training and new loss functions to learn adversarially robust feature representations). Here we offer a unique insight into the predictive behavior of CNNs that they tend to misclassify adversarial samples into the most probable false classes. This inspires us to propose a new Probabilistically Compact (PC) loss with logit constraintsmore »which can be used as a drop-in replacement for cross-entropy (CE) loss to improve CNN’s adversarial robustness. Specifically, PC loss enlarges the probability gaps between true class and false classes meanwhile the logit constraints prevent the gaps from being melted by a small perturbation. We extensively compare our method with the state-of-the-art using large scale datasets under both white-box and black-box attacks to demonstrate its effectiveness. The source codes are available at https://github.com/xinli0928/PC-LC.« less
  7. In this paper, we address the routing and recharging problem for electric vehicles, where charging nodes have heterogeneous prices and waiting times, and the objective is to minimize the total recharging cost. We prove that the problem is NP-hard and propose two algorithms to solve it. The first, is an algorithm which obtains the optimal solution in pseudo-polynomial time. The second, is a polynomial time algorithm that obtains a solution with the total cost of recharging not greater than the optimal cost for a more constrained instance of the problem with the maximum waiting time of (1−ϵ)⋅W , where Wmore »is the maximum allowable waiting time.« less
  8. In mixed-criticality systems, mode switch is a key strategy which dynamically provides a balance between system performance and safety. In conventional MCS frameworks, mode switch is triggered by the over-execution of a task; i.e., a task overruns the less pessimistic worst-case execution time. In cyber-physical systems, the data volume generated by I/O affects and can even dominate task computation time. With this in mind, we introduce a novel MCS architecture, termed Pythia-MCS, which predicts task execution time according to I/O run-time behaviors. With the new feature of future-prediction, the Pythia-MCS provides more timely, but still accurate, mode switch. We alsomore »present a new theoretical model (quarter-clairvoyance), which guarantees the timing predictability of the design, and a new schedulability analysis for the Pythia-MCS, which demonstrates improved schedulability compared to conventional MCS frameworks. The Pythia-MCS is the first MCS framework enabling the clairvoyance functionality.« less
  9. In this paper, we address the Multi-Component Application Placement Problem (MCAPP) in Mobile Edge Computing (MEC) systems. We formulate this problem as a Mixed Integer Non-Linear Program (MINLP) with the objective of minimizing the total cost of running the applications. In our formulation, we take into account two important and challenging characteristics of MEC systems, the mobility of users and the network capabilities. We analyze the complexity of MCAPP and prove that it is NP-hard, that is, finding the optimal solution in reasonable amount of time is infeasible. We design two algorithms, one based on matching and local search andmore »one based on a greedy approach, and evaluate their performance by conducting an extensive experimental analysis driven by two types of user mobility models, real-life mobility traces and random-walk. The results show that the proposed algorithms obtain near-optimal solutions and require small execution times for reasonably large problem instances.« less
  10. A high-voltage-gain dc-dc converter topology is proposed for renewable energy applications. The proposed coupled-inductor-based high-gain dc-dc converter features reduced input current ripple. The semiconductor elements voltage spikes due to the leakage inductance are prevented through the use of a clamping circuit. The Clamping circuit helps recover the leakage inductance stored energy, which causes voltage spikes on the switch. This results in the selection of elements with lower voltage ratings. Power switches with lower voltage ratings lead to lower conduction losses and improved system efficiency. The DC component of the inductor magnetizing current is zero. Consequently, no energy is stored inmore »the inductor core, and the losses are further reduced.« less