检索结果-内蒙古大学图书馆

Local differential privacy in decentralized optimization

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Xiao, Hanshen Ye, Yu Devadas, Srinivas CSAIL MIT School of Electrical Engineering and Computer Science KTH

Privacy concerns with sensitive data are receiving increasing attention. In this paper, we study local differential privacy (LDP) in interactive decentralized optimization. By constructing random local aggregators, we propose a framework to amplify LDP by a constant. We take Alternating Direction Method of Multipliers (ADMM), and decentralized gradient descent as two concrete examples, where experiments support our theory. In an asymptotic view, we address the following question: Under LDP, is it possible to design a distributed private minimizer for arbitrary closed convex constraints with utility loss not explicitly dependent on dimensionality? As an affiliated result, we also show that with merely linear secret sharing, information theoretic privacy is achievable for bounded colluding agents. Copyright © 2019, The Authors. All rights reserved.

关键词： Gradient methods

Zeroth-order hybrid gradient descent: Towards a principled black-box optimization framework

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Sharma, Pranay Xu, Kaidi Liu, Sijia Chen, Pin-Yu Lin, Xue Varshney, Pramod K. Department of Electrical Engineering and Computer Science Syracuse University SyracuseNY13244 United States Department of Computer and Electrical Engineering Northeastern University BostonMA02115 United States MIT-IBM Watson AI Lab IBM Research CambridgeMA02142 United States

In this work, we focus on the study of stochastic zeroth-order (ZO) optimization which does not require first-order gradient information and uses only function evaluations. The problem of ZO optimization has emerged in many recent machine learning applications, where the gradient of the objective function is either unavailable or difficult to compute. In such cases, we can approximate the full gradients or stochastic gradients through function value based gradient estimates. Here, we propose a novel hybrid gradient estimator (HGE), which takes advantage of the query-efficiency of random gradient estimates as well as the variance-reduction of coordinate-wise gradient estimates. We show that with a graceful design in coordinate importance sampling, the proposed HGE-based ZO optimization method is efficient both in terms of iteration complexity as well as function query cost. We provide a thorough theoretical analysis of the convergence of our proposed method for non-convex, convex, and strongly-convex optimization. We show that the convergence rate that we derive generalizes the results for some prominent existing methods in the nonconvex case, and matches the optimal result in the convex case. We also corroborate the theory with a real-world black-box attack generation application to demonstrate the empirical advantage of our method over state-of-the-art ZO optimization approaches. © 2020, CC BY.

关键词： Gradient methods

Low density majority codes and the problem of graceful degradation

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Roozbehani, Hajir Polyanskiy, Yury Laboratory of Information and Decision Systems MIT Department of Electrical Engineering and Computer Science MIT

We study a problem of constructing codes that transform a channel with high bit error rate (BER) into one with low BER (at the expense of rate). Our focus is on obtaining codes with smooth ("graceful") input-output BER curves (as opposed to threshold-like curves typical for long error-correcting codes). This paper restricts attention to binary erasure channels (BEC) and contains three contributions. First, we introduce the notion of Low Density Majority Codes (LDMCs). These codes are non-linear sparse-graph codes, which output majority function evaluated on randomly chosen small subsets of the data bits. This is similar to Low Density Generator Matrix codes (LDGMs), except that the XOR function is replaced with the majority. We show that even with a few iterations of belief propagation (BP) the attained input-output curves provably improve upon performance of any linear systematic code. The effect of non-linearity bootstraping the initial iterations of BP, suggests that LDMCs should improve performance in various applications, where LDGMs have been used traditionally. Second, we establish several two-point converse bounds that lower bound the BER achievable at one erasure probability as a function of BER achieved at another one. The novel nature of our bounds is that they are specific to subclasses of codes (linear systematic and non-linear systematic) and outperform similar bounds implied by the area theorem for the EXIT function. Third, we propose a novel technique for rigorously bounding performance of BP-decoded sparse-graph codes (over arbitrary binary input-symmetric channels). This is based on an extension of ***’s lemma to the χ2-information and a new characterization of the extremality under the less noisy order. Copyright © 2019, The Authors. All rights reserved.

关键词： Belief propagation

Towards an efficient and general framework of robust training for graph neural networks

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Xu, Kaidi Liu, Sijia Chen, Pin-Yu Sun, Mengshu Ding, Caiwen Kailkhura, Bhavya Lin, Xue Electrical & Computer Engineering Northeastern University Boston United States MIT-IBM Watson AI Lab IBM Research Computer Science and Engineering University of Connecticut Storrs United States Lawrence Livermore National Laboratory Livermore United States

Graph Neural Networks (GNNs) have made significant advances on several fundamental inference tasks. As a result, there is a surge of interest in using these models for making potentially important decisions in high-regret applications. However, despite GNNs' impressive performance, it has been observed that carefully crafted perturbations on graph structures (or nodes attributes) lead them to make wrong predictions. Presence of these adversarial examples raises serious security concerns. Most of the existing robust GNN design/training methods are only applicable to white-box settings where model parameters are known and gradient based methods can be used by performing convex relaxation of the discrete graph domain. More importantly, these methods are not efficient and scalable which make them infeasible in time sensitive tasks and massive graph datasets. To overcome these limitations, we propose a general framework which leverages the greedy search algorithms and zeroth-order methods to obtain robust GNNs in a generic and an efficient manner. On several applications, we show that the proposed techniques are significantly less computationally expensive and, in some cases, more robust than the state-of-the-art methods making them suitable to large-scale problems which were out of the reach of traditional robust training methods. Copyright © 2020, The Authors. All rights reserved.

关键词： Graph neural networks

Deep neural network discrimination of multiplexed superconducting qubit states

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Lienhard, Benjamin Vepsäläinen, Antti Govia, Luke C.G. Hoffer, Cole R. Qiu, Jack Y. Ristè, Diego Ware, Matthew Kim, David Winik, Roni Melville, Alexander Niedzielski, Bethany Yoder, Jonilyn Ribeill, Guilhem J. Ohki, Thomas A. Krovi, Hari K. Orlando, Terry P. Gustavsson, Simon Oliver, William D. Department of Electrical Engineering and Computer Science Massachusetts Institute of Technology CambridgeMA02139 United States Research Laboratory of Electronics Massachusetts Institute of Technology CambridgeMA02139 United States Quantum Engineering and Computing Group Raytheon BBN Technologies CambridgeMA02138 United States MIT Lincoln Laboratory LexingtonMA02421 United States

Demonstrating a quantum computational advantage will require high-fidelity control and readout of multi-qubit systems. As system size increases, multiplexed qubit readout becomes a practical necessity to limit the growth of resource overhead. Many contemporary qubit-state discriminators presume single-qubit operating conditions or require considerable computational effort, limiting their potential extensibility. Here, we present multi-qubit readout using neural networks as state discriminators. We compare our approach to contemporary methods employed on a quantum device with five superconducting qubits and frequency-multiplexed readout. We find that fully-connected feedforward neural networks increase the qubit-state-assignment fidelity for our system. Relative to contemporary discriminators, the assignment error rate is reduced by up to 25 % due to the compensation of system-dependent nonidealities such as readout crosstalk which is reduced by up to one order of magnitude. Our work demonstrates a potentially extensible building block for high-fidelity readout relevant to both near-term devices and future fault-tolerant systems. Copyright © 2021, The Authors. All rights reserved.

关键词： Deep neural networks

ARReSVG: Intelligent Multi-UAV Navigation in Partially Observable Spaces Using Adaptive Deep Reinforcement Learning Approach

学校读者我要写书评

暂无评论

IEEE Transactions on Vehicular Technology 2025年

作者： Essaky, Selvam Raja, Gunasekaran Dev, Kapal Niyato, Dusit Anna University MIT Campus NGNLab Department of Computer Technology Chennai600044 India Munster Technological University Department of Computer Science CONNECT Centre Bishopstown CorkT12 P928 Ireland University of Johannesburg Department of Institute of Intelligent Systems Auckland Park2006 South Africa Lebanese American University Department of Electrical and Computer Engineering Byblos1111 Lebanon Nanyang Technological University College of Computing and Data Science Nanyang Avenue 639798 Singapore

Unmanned Aerial Vehicles (UAVs) have brought about a revolution in various applications worldwide. Applications such as aerial surveillance, search and rescue, and operations beyond the line of sight rely on the autonomous navigation capabilities of multi-UAVs. This paper introduces a novel framework, Adaptive Rapid Recurrent Stochastic Valued Gradient (ARReSVG), empowered by Deep Reinforcement Learning (DRL) to address navigational challenges faced by UAVs in partially observable spaces. The proposed ARReSVG framework incorporates an adaptive learning approach to address issues related to partial observability-based autonomous navigation effectively. Compared with traditional obstacle avoidance techniques, the proposed navigation framework utilizes a distributed Multi-UAV Consensus Path Planning (MUCP2) mechanism to perform collaborative path planning for multiple UAVs to ensure collision avoidance trajectories. Additionally, the learned value-action policy implemented in the UAV alleviates the computational demands of real-time map-building. The ARReSVG framework leverages a Gated Recurrent Unit (GRU) to reduce computing complexity and accurately approximate the action-value function. The simulation results show that the ARReSVG framework achieves an impressive 94% © 1967-2012 IEEE.

关键词： Deep reinforcement learning

Hypermultiplexed integrated photonics-based optical tensor processor

学校读者我要写书评

暂无评论

science advances 2025年第23期11卷 eadu0228页

作者： Shaoyuan Ou Kaiwen Xue Lian Zhou Chun-Ho Lee Alexander Sludds Ryan Hamerly Ke Zhang Hanke Feng Yue Yu Reshma Kopparapu Eric Zhong Cheng Wang Dirk Englund Mengjie Yu Zaijun Chen Ming Hsieh Department of Electrical and Computer Engineering University of Southern California Los Angeles CA 90089 USA. Department of Electrical Engineering and Computer Science University of California Berkeley CA 94720 USA. Research Laboratory of Electronics MIT Cambridge MA 02139 USA. PHI Laboratories NTT Research Inc. 940 Stewart Drive Sunnyvale CA 94085 USA. Department of Electrical Engineering & State Key Laboratory of Terahertz and Millimeter Waves City University of Hong Kong Kowloon Hong Kong China. Opticore Inc. Berkeley CA 94705 USA.

The escalating data volume and complexity resulting from the rapid expansion of artificial intelligence (AI), Internet of Things (IoT), and 5G/6G mobile networks is creating an urgent need for energy-efficient, scalable computing hardware. Here, we demonstrate a hypermultiplexed tensor optical processor that can perform trillions of operations per second using space-time-wavelength three-dimensional optical parallelism, enabling O(N) operations per clock cycle with O(N) modulator devices. The system is built with wafer-fabricated III/V micrometer-scale lasers and high-speed thin-film lithium niobate electro-optics for encoding at tens of femtojoules per symbol. Lasing threshold incorporates analog inline rectifier (ReLU) nonlinearity for low-latency activation. The system scalability is verified with machine learning models of 405,000 parameters. A combination of high clock rates, energy-efficient processing, and programmability unlocks the potential of light for low-energy AI accelerators for applications ranging from training of large AI models to real-time decision-making in edge deployment.

关键词：

Variable Frequency Pulse Generation from Breathers in Josephson Transmission Lines

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Cunningham, Gregory Ye, Yufeng Peng, Kaidong Yen, Alec Kedziora, Jessica O’Brien, Kevin P. Harvard John A. Paulson School of Engineering and Applied Sciences Harvard University CambridgeMA02138 United States Research Laboratory of Electronics Massachusetts Institute of Technology CambridgeMA02139 United States Department of Electrical Engineering and Computer Science Massachusetts Institute of Technology CambridgeMA02139 United States MIT Lincoln Laboratory LexingtonMA02421 United States

Single flux quantum technology has the potential to enhance readout and control of superconducting quantum systems due to their low energy consumption, high speed, and cryogenic operating temperatures. Current cryogenic readout and control typically requires microwave pulses of specific frequencies to travel between the room temperature control electronics and the cryogenic setup. Latency in control and readout can be improved by generating pulses within the dilution refrigerator. In this work, we consider a protocol for generating gigahertz frequency microwave tones from trains of DC-centered fluxons and fluxoids in Josephson transmission lines using the dynamics of breather formation, without room temperature synthesis or shunt / bias resistors. Simulations show that pulses with frequencies in the range of 15.2 to 21.5 GHz can be generated with maximal energy efficiency of 97% and bandwidth from 40 to 365 MHz. This protocol can also be used to generate gigahertz frequency Gaussian pulses. We detail metrics relevant to the control and readout of quantum systems such as input power, output power, and footprint. Copyright © 2025, The Authors. All rights reserved.

关键词： Josephson junction devices

Cyber-Physical Security and Resiliency Analysis Testbed for Critical Microgrids with IEEE 2030.5

学校读者我要写书评

暂无评论

Cyber-Physical Security and Resiliency Analysis Testbed for ...

Workshop Modeling and Simulation of Cyber-Physical Energy Systems (MSCPES)

作者： Partha S. Sarker V. Venkataramanan D. Sebastian Cardenas A. Srivastava A. Hahn B. Miller School of Electrical Engineering and Computer Science Washington State University Pullman WA 99163 USA currently a post-doctoral research associate with MIT National Renewable Energy Lab (NREL)

ISBN: (数字)9781728187211

ISBN: (纸本)9781728187228

The development of more resilient grids is an on-going effort that has attracted multiple participants. Within this context, Distributed Energy Resources, along with transactive energy mechanisms are being considered as the key driving technologies. Yet their under-laying communication capabilities might introduce additional cybersecurity risks that must be analysed. This paper proposes a cyber-physical microgrid testbed using OpenDSS, Mininet and IEEE 2030.5 that can be used to study the grid's cyber-resilience under various scenarios. For critical microgrid installations, it is essential that the critical loads are served in spite of multiple contingencies. A resiliency analysis is proposed for a military microgrid to study its performance with these contingencies and the results are analyzed.

关键词：