检索结果-内蒙古大学图书馆

Learning to Coordinate in Multi-Agent systems: A Coordinated Actor-Critic Algorithm and Finite-Time Guarantees

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zeng, Siliang Chen, Tianyi Garcia, Alfredo Hong, Mingyi Department of Electrical and Computer Engineering University of Minnesota MN United States Department of Electrical Computer and Systems Engineering Rensselaer Polytechnic Institute NY United States Department of Industrial and Systems Engineering Texas A&M University TX United States

Multi-agent reinforcement learning (MARL) has attracted much research attention recently. However, unlike its single-agent counterpart, many theoretical and algorithmic aspects of MARL have not been well-understood. In this paper, we study the emergence of coordinated behavior by autonomous agents using an actor-critic (AC) algorithm. Specifically, we propose and analyze a class of coordinated actor-critic (CAC) algorithms in which individually parametrized policies have a shared part (which is jointly optimized among all agents) and a personalized part (which is only locally optimized). Such a kind of partially personalized policy allows agents to coordinate by leveraging peers’ experience and adapt to individual tasks. The flexibility in our design allows the proposed CAC algorithm to be used in a fully decentralized setting, where the agents can only communicate with their neighbors, as well as in a federated setting, where the agents occasionally communicate with a server while optimizing their (partially personalized) local models. Theoretically, we show that under some standard regularity assumptions, the proposed CAC algorithm requires O(Ε-52) samples to achieve an Ε-stationary solution (defined as the solution whose squared norm of the gradient of the objective function is less than Ε). To the best of our knowledge, this work provides the first finite-sample guarantee for decentralized AC algorithm with partially personalized policies. Copyright © 2021, The Authors. All rights reserved.

关键词： Autonomous agents

Effects of control volume outlet variation on axial air cooling of lithium-ion batteries

学校读者我要写书评

暂无评论

International Journal of Thermofluids 2023年 19卷

作者： Mahek, Mehwish Khan Alkhedher, Mohammad Ghazal, Mohammed Abdelkareem, Mohammad Ali Ramadan, Mohamad Olabi, Abdul-Ghani Mechanical and Industrial Engineering Department College of Engineering Abu Dhabi University United Arab Emirates Electrical Computer and Biomedical Engineering Department College of Engineering Abu Dhabi University United Arab Emirates Sustainable Energy & Power Systems Research Centre RISE University of Sharjah Sharjah P.O. Box 27272 United Arab Emirates Chemical Engineering Department Minia University Elminia Egypt School of Engineering International University of Beirut BIU Beirut Lebanon School of Engineering Lebanese International University LIU Beirut Lebanon Mechanical Engineering and Design Aston University School of Engineering and Applied Science Aston Triangle BirminghamB4 7ET United Kingdom

This paper addresses cylindrical lithium-ion cells' heating and non-uniform cooling by simulating a simple battery pack design with 24 cells and axial airflow. Cooling a battery pack using forced air convection in a rectangular container causes non-homogeneous temperature distribution. The middle cells and the second half of the battery pack (near the exit) get less cooling compared to the cells near the inlet. To solve this issue of non-uniform cooling, in this paper, we propose an optimized Battery Thermal Management System (BTMS) which allows higher amounts of turbulent kinetic energy to be generated near the cells closer to the exit and farther away from the inlet. We attempted to create a narrow region in the middle of the pack which leads to an increase in mixing and turbulence in the airflow. This directs the flow precisely, hence getting rid of dead air flow zones behind the cells. The convergence functions like a nozzle, propelling the air flow into the second half of the battery pack. Using COMSOL Multiphysics, the mixing and turbulence of the airflow, as well as the temperature increase and homogeneity inside the battery pack, are investigated. Three separate investigations are conducted in order to determine the optimal shape for the BTMS control volume as well as to determine the effect of system's inlet pressure and outlet velocity on the cooling performance. Results indicated that the system's temperature decreased significantly by lowering the outlet dimensions and introducing convergence to the BTMS's central area. Increasing outlet velocity increased cooling in a case with a single fan significantly, whereas increasing pressure in a case with two fans increased cooling slightly. © 2023 The Authors

关键词： Lithium-ion batteries

Classification vs regression in overparameterized regimes: does the loss function matter?

学校读者我要写书评

暂无评论

The Journal of Machine Learning Research

The Journal of Machine Learning Research 2021年第1期22卷 10104-10172页

作者： Vidya Muthukumar Adhyyan Narang Vignesh Subramanian Mikhail Belkin Daniel Hsu Anant Sahai Electrical and Computer Engineering and Industrial and Systems Engineering Georgia Institute of Technology Atlanta GA Department of Electrical and Computer Engineering University of Washington Seattle WA Department of Electrical Engineering and Computer Sciences University of California Berkeley Berkeley CA Halicioğlu Data Science Institute UC San Diego La Jolla CA Department of Computer Science and Data Science Institute Columbia University New York NY

We compare classification and regression tasks in an overparameterized linear model with Gaussian features. On the one hand, we show that with sufficient overparameterization all training points are support vectors: solutions obtained by least-squares minimum-norm interpolation, typically used for regression, are identical to those produced by the hard-margin support vector machine (SVM) that minimizes the hinge loss, typically used for training classifiers. On the other hand, we show that there exist regimes where these interpolating solutions generalize well when evaluated by the 0-1 test loss function, but do not generalize if evaluated by the square loss function, i.e. they approach the null risk. Our results demonstrate the very different roles and properties of loss functions used at the training phase (optimization) and the testing phase (generalization).

关键词： classification regression overparameterized support vector machines survival contamination

Smart Fibers for Energy Transformation and Storing

学校读者我要写书评

暂无评论

Smart Fibers for Energy Transformation and Storing

Advances in Science and engineering Technology International Conferences (ASET)

作者： Haya Rami Al Jaghoub Maryam Nooman AlMallahi Khaled Obaideen Asmaa Moustafa Wafaa Al Nassan Omar H. Ramadan Talal Bonny Industrial Engineering and Engineering Management University of Sharjah Sharjah United Arab Emirates Department of Mechanical and Aerospace Engineering United Arab Emirates University Al Ain United Arab Emirates Sustainable Energy & Power Systems Research Centre RISE University of Sharjah Sharjah United Arab Emirates Faculty of Engineering Ahram Canadian University Giza Egypt Department of Computer Engineering University of Sharjah Sharjah United Arab Emirates Department of Electrical and Electronic Engineering University of Sharjah Sharjah United Arab Emirates

Fibers have been a crucial element in the development of textiles. This is attributed to the numerous advantages fibers provide, including the production of stronger and long-lasting cloths. Moreover, synthetic fibers are perceived as highly flexible and cost-effective materials. Due to the expansion in portable and wearable devices, the creation of intelligent and versatile fibers has been significantly boosted. Additionally, due to the intrinsic materials of fibers, such as semiconductors and electroactive materials, it is feasible to create fiber electronics. As a result, this paper focuses on fiber electronics and especially the sources of energy supply for these fibers. This paper initially starts to introduce the development of functional fibers in energy conversion and storage, with an emphasis on supercapacitors, solar cells, and nanogenerator batteries. Thereafter, the significance of fiber-shaped energy conversion and smart structural design for storage devices is examined.

关键词：

Conditional Gradient Methods

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Braun, Gábor Carderera, Alejandro Combettes, Cyrille W. Hassani, Hamed Karbasi, Amin Mokhtari, Aryan Pokutta, Sebastian Department for AI in Society Science and Technology Zuse Institute Berlin Germany Institute of Mathematics Technische Universität Berlin Germany School of Industrial and Systems Engineering Georgia Institute of Technology United States Toulouse School of Economics Université Toulouse 1 Capitole France Department of Electrical and Systems Engineering University of Pennsylvania United States Department of Electrical Engineering Computer Science Statistics & Data Science Yale University United States Department of Electrical and Computer Engineering University of Texas at Austin United States

The purpose of this survey is to serve both as a gentle introduction and acoherent overview of state-of-the-art Frank-Wolfe algorithms, also calledconditional gradient algorithms, for function minimization. These algorithmsare especially useful in convex optimization when linear optimization ischeaper than projections. The selection of the material has been guided by the principle ofhighlighting crucial ideas as well as presenting new approaches that we believemight become important in the future, with ample citations even of old worksimperative in the development of newer methods. Yet, our selection is sometimesbiased, and need not reflect consensus of the research community, and we havecertainly missed recent important contributions. After all the research area ofFrank-Wolfe is very active, making it a moving target. We apologize sincerelyin advance for any such distortions and we fully acknowledge: We stand on theshoulder of giants. Copyright © 2022, The Authors. All rights reserved.

关键词： Convex optimization

Improving neural network robustness via persistency of excitation

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Sridhar, Kaustubh Sokolsky, Oleg Lee, Insup Weimer, James The PRECISE Center The Departments of Electrical & Systems Engineering and Computer & Information Science University of Pennsylvania PhiladelphiaPA United States

Improving adversarial robustness of neural networks remains a major challenge. Fundamentally, training a neural network via gradient descent is a parameter estimation problem. In adaptive control, maintaining persistency of excitation (PoE) is integral to ensuring convergence of parameter estimates in dynamical systems to their true values. We show that parameter estimation with gradient descent can be modeled as a sampling of an adaptive linear time-varying continuous system. Leveraging this model, and with inspiration from Model-Reference Adaptive Control (MRAC), we prove a sufficient condition to constrain gradient descent updates to reference persistently excited trajectories converging to the true parameters. The sufficient condition is achieved when the learning rate is less than the inverse of the Lipschitz constant of the gradient of loss function. We provide an efficient technique for estimating the corresponding Lipschitz constant in practice using extreme value theory. Our experimental results in both standard and adversarial training illustrate that networks trained with the PoE-motivated learning rate schedule have similar clean accuracy but are significantly more robust to adversarial attacks than models trained using current state-of-the-art heuristics. Copyright © 2021, The Authors. All rights reserved.

关键词： Gradient methods

MeWEHV: Mel and Wave Embeddings for Human Voice Tasks

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Carofilis, Andrés Fernández-Robles, Laura Alegre, Enrique Fidalgo, Eduardo Dept. of Electrical Systems and Automation Engineering School of Industrial Computer and Aerospace Engineering Universidad de León León24007 Spain Researcher at INCIBE Spanish National Cybersecurity Institute León Spain Dept. of Mechanical Computer and Aerospace Engineering Universidad de León Spain

A recent trend in speech processing is the use of embeddings created through machine learning models trained on a specific task with large datasets. By leveraging the knowledge already acquired, these models can be reused in new tasks where the amount of available data is small. This paper proposes a pipeline to create a new model, called Mel and Wave Embeddings for Human Voice Tasks (MeWEHV), capable of generating robust embeddings for speech processing. MeWEHV combines the embeddings generated by a pre-trained raw audio waveform encoder model, and deep features extracted from Mel Frequency Cepstral Coefficients (MFCCs) using Convolutional Neural Networks (CNNs). We evaluate the performance of MeWEHV on three tasks: speaker, language, and accent identification. For the first one, we use the VoxCeleb1 dataset and present YouSpeakers204, a new and publicly available dataset for English speaker identification that contains 19607 audio clips from 204 persons speaking in six different accents, allowing other researchers to work with a very balanced dataset, and to create new models that are robust to multiple accents. For evaluating the language identification task, we use the VoxForge and Common Language datasets. Finally, for accent identification, we use the Latin American Spanish Corpora (LASC) and Common Voice datasets. Our approach allows a significant increase in the performance of state-of-the-art models on all the tested datasets, with a low additional computational cost. © 2022, CC BY-NC-ND.

关键词： Embeddings

Harmless interpolation in regression and classification with structured features

学校读者我要写书评

暂无评论

arXiv 2021年

作者： McRae, Andrew D. Karnik, Santhosh Davenport, Mark A. Muthukumar, Vidya School of Electrical and Computer Engineering Georgia Tech Department of Computational Mathematics Science & Engineering Michigan State University School of Industrial and Systems Engineering Georgia Tech

Overparametrized neural networks tend to perfectly fit noisy training data yet generalize well on test data. Inspired by this empirical observation, recent work has sought to understand this phenomenon of benign overfitting or harmless interpolation in the much simpler linear model. Previous theoretical work critically assumes that either the data features are statistically independent or the input data is high-dimensional;this precludes general nonparametric settings with structured feature maps. In this paper, we present a general and flexible framework for upper bounding regression and classification risk in a reproducing kernel Hilbert space. A key contribution is that our framework describes precise sufficient conditions on the data Gram matrix under which harmless interpolation occurs. Our results recover prior independent-features results (with a much simpler analysis), but they furthermore show that harmless interpolation can occur in more general settings such as features that are a bounded orthonormal system. Furthermore, our results show an asymptotic separation between classification and regression performance in a manner that was previously only shown for Gaussian features. Copyright © 2021, The Authors. All rights reserved.

关键词： Interpolation

The good, the bad and the ugly sides of data augmentation: An implicit spectral regularization perspective

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Lin, Chi-Heng Kaushik, Chiraag Dyer, Eva L. Muthukumar, Vidya School of Electrical and Computer Engineering Georgia Institute of Technology GA United States Department of Biomedical Engineering Georgia Institute of Technology GA United States School of Industrial and Systems Engineering Georgia Institute of Technology GA United States Samsung Research America United States

Data augmentation (DA) is a powerful workhorse for bolstering performance in modern machine learning. Specific augmentations like translations and scaling in computer vision are traditionally believed to improve generalization by generating new (artificial) data from the same distribution. However, this traditional viewpoint does not explain the success of prevalent augmentations in modern machine learning (e.g. randomized masking, cutout, mixup), that greatly alter the training data distribution. In this work, we develop a new theoretical framework to characterize the impact of a general class of DA on underparameterized and overparameterized linear model generalization. Our framework reveals that DA induces implicit spectral regularization through a combination of two distinct effects: a) manipulating the relative proportion of eigenvalues of the data covariance matrix in a training-data-dependent manner, and b) uniformly boosting the entire spectrum of the data covariance matrix through ridge regression. These effects, when applied to popular augmentations, give rise to a wide variety of phenomena, including discrepancies in generalization between over-parameterized and under-parameterized regimes and differences between regression and classification tasks. Our framework highlights the nuanced and sometimes surprising impacts of DA on generalization, and serves as a testbed for novel augmentation design. Copyright © 2022, The Authors. All rights reserved.

关键词： Machine learning