检索结果-内蒙古大学图书馆

Decentralized Personalized Federated Learning for Min-Max Problems

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Borodich, Ekaterina Beznosikov, Aleksandr Sadiev, Abdurakhmon Sushko, Vadim Savelyev, Nikolay Takac, Martin Gasnikov, Alexander Moscow Institute of Physics and Technology Moscow Russia Mohamed bin Zayed University of Artificial Intelligence Abu Dhabi United Arab Emirates Ivannikov Institute for System Programming RAS Moscow Russia Bosch Center for Artificial Intelligence Renningen Germany

Personalized Federated Learning (PFL) has witnessed remarkable advancements, enabling the development of innovative machine learning applications that preserve the privacy of training data. However, existing theoretical research in this field has primarily focused on distributed optimization for minimization problems. This paper is the first to study PFL for saddle point problems encompassing a broader range of optimization problems, that require more than just solving minimization problems. In this work, we consider a recently proposed PFL setting with the mixing objective function, an approach combining the learning of a global model together with locally distributed learners. Unlike most previous work, which considered only the centralized setting, we work in a more general and decentralized setup that allows us to design and analyze more practical and federated ways to connect devices to the network. We proposed new algorithms to address this problem and provide a theoretical analysis of the smooth (strongly) convex–(strongly) concave saddle point problems in stochastic and deterministic cases. Numerical experiments for bilinear problems and neural networks with adversarial noise demonstrate the effectiveness of the proposed methods. Copyright © 2021, The Authors. All rights reserved.

关键词： Numerical methods

Ti-Patch: Tiled Physical Adversarial Patch for no-reference video quality metrics

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Leonenkova, Victoria Shumitskaya, Ekaterina Antsiferova, Anastasia Vatolin, Dmitriy Lomonosov Moscow State University Leninskiye Gory 1 Moscow119991 Russia ISP RAS Research Center for Trusted Artificial Intelligence Ivannikov Institute for System Programming of the RAS Alexander Solzhenitsyn st. 25 Moscow109004 Russia MSU Institute for Artificial Intelligence Lomonosovskiy Prospekt 27/1 Moscow119192 Russia

Objective no-reference image- and video-quality metrics are crucial in many computer vision tasks. However, state-of-the-art no-reference metrics have become learning-based and are vulnerable to adversarial attacks. The vulnerability of quality metrics imposes restrictions on using such metrics in quality control systems and comparing objective algorithms. Also, using vulnerable metrics as a loss for deep learning model training can mislead training to worsen visual quality. Because of that, quality metrics testing for vulnerability is a task of current interest. This paper proposes a new method for testing quality metrics vulnerability in the physical space. To our knowledge, quality metrics were not previously tested for vulnerability to this attack;they were only tested in the pixel space. We applied a physical adversarial Ti-Patch – Tiled Patch – attack to quality metrics and did experiments both in pixel and physical space. We also performed experiments on the implementation of physical adversarial wallpaper. The proposed method can be used as additional quality metrics in vulnerability evaluation, complementing traditional subjective comparison and vulnerability tests in the pixel space. We made our code and adversarial videos available on GitHub: https://***/leonenkova/Ti-Patch. © 2024, CC BY.

关键词： Pixels

Evaluation of interpretability methods for adversarial robustness on real-world datasets

学校读者我要写书评

暂无评论

Evaluation of interpretability methods for adversarial robus...

Ivannikov Memorial Workshop (IVMEM)

作者： Anna Chistyakova Maria Cherepnina Konstantin Arkhipenko Sergey D. Kuznetsov Chang-Seok Oh Sebeom Park Lomonosov Moscow State University Moscow Russia Ivannikov Institute for System Programming of the RAS Moscow Russia Moscow Institute of Physics and Technology (State University) Moscow Russia Plekhanov Russian University of Economics Moscow Russia Samsung Research Seoul South Korea

ISBN: (纸本)9781665423281

Adversarial training is considered the most powerful approach for robustness against attacks on deep neural networks involving adversarial examples. However, recent works have shown that the similar robustness level can be achieved by other means, namely interpretability-based regularization. We evaluate these interpretability-based approaches on real-world ResNet models trained on CIFAR-10 and ImageNet datasets. Our results show that interpretability can marginally improve robustness when combined with adversarial training, however, they bring additional computational complexity making these approaches questionable for such models and datasets.

关键词： Training Deep learning Computational modeling Conferences Neural networks Robustness Computational complexity

The first optimal algorithm for smooth and strongly-convex-strongly-concave minimax optimization 22

学校读者我要写书评

暂无评论

The first optimal algorithm for smooth and strongly-convex-s...

Proceedings of the 36th International Conference on Neural Information Processing systems

作者： Dmitry Kovalev Alexander Gasnikov King Abdullah University of Science and Technology Thuwal Saudi Arabia Moscow Institute of Physics and Technology Dolgoprudny Russia and Institute for System Programming RAS Research Center for Trusted Artificial Intelligence Moscow Russia and National Research University Higher School of Economics Moscow Russia

ISBN: (纸本)9781713871088

In this paper, we revisit the smooth and strongly-convex-strongly-concave minimax optimization problem. Zhang et al. (2021) and Ibrahim et al. (2020) established the lower bound $\Omega\left(\sqrt{\kappa_x\kappa_y} \log \frac{1}{\epsilon}\right)$ on the number of gradient evaluations required to find an ε-accurate solution, where kx and ky are condition numbers for the strong convexity and strong concavity assumptions. However, the existing state-of-the-art methods do not match this lower bound: algorithms of Lin et al. (2020) and Wang and Li (2020) have gradient evaluation complexity $\mathcal{O}\left(\sqrt{\kappa_x\kappa_y} \log^3 \frac{1}{\epsilon}\right)$ and $\mathcal{O}\left( \sqrt{\kappa_x\kappa_y}\log^3 (\kappa_x\kappa_y)\log\frac{1}{\epsilon}\right)$, respectively. We fix this fundamental issue by providing the first algorithm with O (√kxky log 1/ε) gradient evaluation complexity. We design our algorithm in three steps: (i) we reformulate the original problem as a minimization problem via the pointwise conjugate function; (ii) we apply a specific variant of the proximal point algorithm to the reformulated problem; (iii) we compute the proximal operator inexactly using the optimal algorithm for operator norm reduction in monotone inclusions.

关键词：

Accelerated primal-dual gradient method for smooth and convex-concave saddle-point problems with bilinear coupling 22

学校读者我要写书评

暂无评论

Accelerated primal-dual gradient method for smooth and conve...

Proceedings of the 36th International Conference on Neural Information Processing systems

作者： Dmitry Kovalev Alexander Gasnikov Peter Richtárik King Abdullah University of Science and Technology Thuwal Saudi Arabia Moscow Institute of Physics and Technology Dolgoprudny Russia and Institute for System Programming RAS Research Center for Trusted Artificial Intelligence Moscow Russia and National Research University Higher School of Economics Moscow Russia

ISBN: (纸本)9781713871088

In this paper we study the convex-concave saddle-point problem minx maxy f(x) + yTAx – g(y), where f(x) and g(y) are smooth and convex functions. We propose an Accelerated Primal-Dual Gradient Method (APDG) for solving this problem, achieving (i) an optimal linear convergence rate in the strongly-convex-strongly-concave regime, matching the lower complexity bound (Zhang et al., 2021), and (ii) an accelerated linear convergence rate in the case when only one of the functions f(x) and g(y) is strongly convex or even none of them are. Finally, we obtain a linearly convergent algorithm for the general smooth and convex-concave saddle point problem minx maxy F(x, y) without the requirement of strong convexity or strong concavity.

关键词：

3D hydrodynamical simulation of accretion disk in binary star system using RKDG CFD solver

学校读者我要写书评

暂无评论

Journal of Physics: Conference Series 2018年第1期1103卷

作者： M P Galanin V V Lukin V M Chechetkin Keldysh Institute of Applied Mathematics of the RAS Moscow Russia Ivannikov Institute for System Programming of the RAS Moscow Russia

We present the results of 3D-hydrodynamical simulations of accretion disk in close binary star system. The model includes the optical star filling its Roche lobe, a gas stream emanating from the inner Lagrangian point of the binary system, and the accretion disc structure. A cold hydrogen gas stream is initially emitted towards a point-like gravitational centre. A stationary accretion disc is formed in about 5 orbital periods after the beginning of accretion. The model uses realistic cooling function for hydrogen. The simulation of accreting gas dynamics is performed using second order Runge — Kutta discontinuous Galerkin CFD solver for unstructured tetrahedral meshes. The monotonicity technique based on Hermite WENO solution reconstruction is implemented. Utilizing the hydrodynamical simulations the sinthetic light curves of the system are calculated as the volume emission of optically thin layers along the line of sight. The simulations results are in a good agreement with observations.

关键词：

A Simple and Effective Method of Cross-Lingual Plagiarism Detection

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Avetisyan, Karen Malajyan, Arthur Ghukasyan, Tsolak Avetisyan, Arutyun Institute for System Programming The Russian Academy of Sciences Alexander Solzhenitsyn 25 Moscow109004 Russia Russian-Armenian University Hovsep Emin 123 Yerevan0051 Armenia ISP RAS Research Center for Trusted Artificial Intelligence Alexander Solzhenitsyn 25 Moscow109004 Russia

We present a simple cross-lingual plagiarism detection method applicable to a large number of languages. The presented approach leverages open multilingual thesauri for candidate retrieval task and pre-trained multilingual BERT-based language models for detailed analysis. The method does not rely on machine translation and word sense disambiguation when in use, and therefore is suitable for a large number of languages, including under-resourced languages. The effectiveness of the proposed approach is demonstrated for several existing and new benchmarks, achieving state-of-the-art results for French, Russian, and Armenian languages. Copyright © 2023, The Authors. All rights reserved.

关键词： Thesauri

Russian DATA LAKE PROTOTYPE AS AN APPROACH towards NATIONAL FEDERATED STORAGE for MEGASCIENCE 9

学校读者我要写书评

暂无评论

Russian DATA LAKE PROTOTYPE AS AN APPROACH towards NATIONAL ...

9th International Conference "Distributed Computing and Grid Technologies in Science and Education", GRID 2021

作者： Alekseev, Aleksandr Kiryanov, Andrey Klimentov, Aleksey Korchuganova, Tatyana Oleynik, Danila Zarochentsev, Anderey NRC "Kurchatov Institute" 1 Akademika Kurchatova sq. Moscow123182 Russia Petersburg Nuclear Physics Institute of NRC "Kurchatov Institute" 1 Orlova Rocha Gatchina 188300 Russia Saint Petersburg State University 7-9 Universitetskaya emb. Saint Petersburg199034 Russia Brookhaven National Laboratory UptonNY United States Ivannikov Institute for System Programming RAS 25 Alexander Solzhenitsyn st. Moscow109004 Russia University Andres Bello Santiago Chile Plekhanov Russian University of Economics 36 Stremyanny lane Moscow117997 Russia Joint Institute for Nuclear Research 6 Joliot-Curie st. Dubna 141980 Russia

A substantial data volume growth will appear with the start of the HL-LHC era. It is not well covered by the current LHC computing model, even taking into account the hardware evolution. The WLCG DOMA project was established to provide data management and storage researches. National data lake r&d's, as a part of the DOMA project, should address the study of possible technology solutions for the organization of intelligent distributed federated storage. This talk will present the current status of the Russian Scientific Data lake prototype and the methodology, which is used for the validation and functional testing of deployed infrastructure. Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).

关键词： Lakes

QUANTUM RANDOM WALKS AND QUANTUM OSCILLATOR IN AN INFINITE-DIMENSIONAL PHASE SPACE

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Busovikov, Vladimir Pechen, Alexander Sakbaev, Vsevolod Department of Mathematical Methods for Quantum Technologies Steklov Mathematical Institute of Russian Academy of Sciences Gubkina Str. 8 Moscow Russia University of Science and Technology MISIS Leninskiy prosp. 4 Moscow Russia Ivannikov Institute for System Programming of RAS Alexander Solzhenitsyn Str. 25 Moscow Russia

We consider quantum random walks in an infinite-dimensional phase space constructed using Weyl representation of the coordinate and momentum operators in the space of functions on a Hilbert space which are square integrable with respect to a shift-invariant measure. We study unitary groups of shift operators in the phase space and averaging of such shifts by Gaussian vectors, which form semigroups of self-adjoint contractions: we find conditions for their strong continuity and establish properties of their generators. Significant differences in their properties allow us to show the absence of the Fourier transform as a unitary transformation that implements the unitary equivalence of these compressive semigroups. Next, we prove the Taylor formula for a certain special subset of smooth functions for shifting to a non-finite vector. It allows us to prove convergence of quantum random walks in the coordinate representation to the evolution of a diffusion process, as well as convergence of quantum random walks in both coordinate and momentum representations to the evolution semigroup of a quantum oscillator in an infinite-dimensional phase space. We find the special essential common domain of generators of semigroups arising in averaging of random shift operators both in position and momentum representations. The invariance of this common domain with respect to both semigroups allows to establish properties of a convex combination of both generators. That convex combination are Hamiltonians of infinite-dimensional quantum oscillators. Thus, we obtain that a Weyl representation of a random walk in an infinite dimensional phase space describes the semigroup of self-adjoint contractions whose generator is the Hamiltonian of an infinite dimensional harmonic *** Codes 28C20, 81V73, 43A32, 37K06 Copyright © 2024, The Authors. All rights reserved.

关键词： Hilbert spaces