Filters








27 Hits in 5.5 sec

Revisiting Hilbert-Schmidt Information Bottleneck for Adversarial Robustness [article]

Zifeng Wang, Tong Jian, Aria Masoomi, Stratis Ioannidis, Jennifer Dy
2021 arXiv   pre-print
We investigate the HSIC (Hilbert-Schmidt independence criterion) bottleneck as a regularizer for learning an adversarially robust deep neural network classifier.  ...  We show that the HSIC bottleneck enhances robustness to adversarial attacks both theoretically and experimentally.  ...  We revisit the HSIC bottleneck, studying its adversarial robustness properties. In contrast to both Alemi et al. [1] and Ma et al.  ... 
arXiv:2106.02734v2 fatcat:q4b45ttwljhzxg45qgfpnr3j7i

Stronger and Faster Wasserstein Adversarial Attacks [article]

Kaiwen Wu and Allen Houze Wang and Yaoliang Yu
2020 arXiv   pre-print
Furthermore, employing our stronger attacks in adversarial training significantly improves the robustness of adversarially trained models.  ...  measuring image quality and has recently risen as a compelling alternative to the ℓ_p metric in adversarial attacks.  ...  Semidef- inite relaxations for certifying robustness to adversarial examples. In Advances in Neural Information Processing Systems 31, pp. 10877-10887, 2018. Rubner, Y., Guibas, L.  ... 
arXiv:2008.02883v1 fatcat:xeu7cw3ctrfopit2dgw662gqe4

Learning Disentangled Representations in the Imaging Domain [article]

Xiao Liu, Pedro Sanchez, Spyridon Thermos, Alison Q. O'Neil, Sotirios A. Tsaftaris
2022 arXiv   pre-print
In this tutorial paper, we motivate the need for disentangled representations, revisit key concepts, and describe practical building blocks and criteria for learning such representations.  ...  This alleviation of the data and annotation requirements offers tantalising prospects for applications in computer vision and healthcare.  ...  We thank the participants of the DREAM tutorials for feedback.  ... 
arXiv:2108.12043v5 fatcat:cbpmp6pbajhjvjzovulswuj2wy

Value Function Approximations via Kernel Embeddings for No-Regret Reinforcement Learning [article]

Sayak Ray Chowdhury, Rafael Oliveira
2022 arXiv   pre-print
In this paper, we propose an online model-based RL algorithm, namely the CME-RL, that learns representations of transition distributions as embeddings in a reproducing kernel Hilbert space while carefully  ...  However, the understanding of function approximation schemes for state-value functions largely remains missing.  ...  However, observe that for any operator M ∈ L(H ϕ , H ψ ) we have M tr(M ⊤ M), where the latter corresponds to the Hilbert-Schmidt norm.  ... 
arXiv:2011.07881v3 fatcat:qvedjipt4ndlvlq4fe5rzdcoqq

Solving inverse problems using data-driven models

Simon Arridge, Peter Maass, Ozan Öktem, Carola-Bibiane Schönlieb
2019 Acta Numerica  
Recent research in inverse problems seeks to develop a mathematically coherent foundation for combining data-driven models, and in particular those based on deep learning, with domain-specific knowledge  ...  The authors are moreover grateful to the following people for proofreading the manuscript and providing valuable feedback on its content  ...  Clearly, theory for robustness against adversarial attacks in the context of inverse problems is very much an emerging field.  ... 
doi:10.1017/s0962492919000059 fatcat:2f7te542wrftphdhurcdnw6dqu

Advances in Quantum Cryptography

Stefano Pirandola, Ulrik Andersen, Leonardo Banchi, Mario Berta, Darius Bunandar, Roger Colbeck, Dirk Englund, Tobias Gehring, Cosmo Lupo, Carlo Ottaviani, Jason Pereira, Mohsen Razavi (+6 others)
2020 Advances in Optics and Photonics  
General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications  ...  that users recognise and abide by the legal requirements associated with these rights.  Users may download and print one copy of any publication from the public portal for the purpose of private study  ...  The CHSH inequality can be expressed in this notation as C, P ≤ 2, where P = P z XY |AB , C =      1 −1 1 −1 −1 1 −1 1 1 −1 −1 1 −1 1 1 −1      and C, P = Tr(C T P ) is the Hilbert-Schmidt inner  ... 
doi:10.1364/aop.361502 fatcat:7xp7i42fjffpblk6g4fhlpgg6m

The Modern Mathematics of Deep Learning [article]

Julius Berner, Philipp Grohs, Gitta Kutyniok, Philipp Petersen
2021 arXiv   pre-print
For selected approaches, we describe the main ideas in more detail.  ...  eigenvalues of the Hessian (or approximately as the robustness of the minimizer to adversarial perturbations on the parameter space), see [KMN+ 17].  ...  We will revisit this point in Assumption 1.10 and Definition 1.11.  ... 
arXiv:2105.04026v1 fatcat:lxnfyzr6qfasneo433inpgseia

Self-Organized Evolutionary Process in Sets of Interdependent Variables near the Midpoint of Phase Transition in K-Satisfiability [chapter]

Michael Korkin
2001 Lecture Notes in Computer Science  
Ackerman, John T Strategic Studies Quarterly: An Air Force-Sponsored Strategic Forum for Military, Government, and Academic Professionals. Volume 2, Number 1 -7 Acuff, Hugh F  ...  The requirement for larger memory while building big databases can sometimes make this resource a bottleneck for an information indexing system.  ...  Hilbert spaces.  ... 
doi:10.1007/3-540-45443-8_20 fatcat:ght26tu6avh57kkp54kifofaee

Cross-Covariance Models [chapter]

2017 Encyclopedia of GIS  
Synonyms Cadaster; Land administration system; Land information system; Land policy; Land registry; Property register; Spatial reference frames  ...  Cross-References Indexing, Hilbert R-tree, Spatial Indexing, Multimedia Indexing The explosive growth of spatial data and widespread use of spatial databases emphasize the need for the automated discovery  ...  Efficient tools for extracting information from geo-spatial data are crucial to organizations which make decisions based on large spatial datasets, including the National Aeronautics and Space Administration  ... 
doi:10.1007/978-3-319-17885-1_100240 fatcat:2ojzb7es7rhofinw4abol6dgc4

Abstracts of Working Papers in Economics

2000 Abstracts of Working Papers in Economics  
Each PRP's type is its private information about the precision of its own records relating to the site. A strategy for a PRP is a function mapping its type into announced levels of precision.  ...  The authors model the problem as an incomplete information, simultaneous-move game between PRP's.  ...  Rho is a linear operator on a Hilbert space. X subscript t and epsilon subscript t are Hilbert space valued random variables (the epsilon-subscript-t's are iid).  ... 
doi:10.1017/s0951007900004915 fatcat:62y3byrmc5ag5czgxtoyqmhhdm

Machine learning approaches for time series problems

Χριστόφορος Στ. Ναλμπάντης
2022
The first pillar of our scientific contributions comprises novel machine learning solutions for the problem of NILM.  ...  These aspects bring out many new challenges when trying to extract useful information from time series data using machine learning methods.  ...  Grigorios Tsoumakas for their scientific knowledge, their honest feedback and career advice that they provided.  ... 
doi:10.26262/heal.auth.ir.340363 fatcat:fqggq54crnaxpmaixk4go7kraa

DISSERTATION SEARCHING OVER ENCRYPTED DATA Submitted by

Tarik Moataz, Frédéric Cuppens, Nora Boulahia, Cuppens Wang, Haonan Mcconnell
2017 unpublished
, bank account information, social security numbers, data income and more information for more than millions of customers and users.  ...  Concerns related to data confidentiality rise uncertainty for users maintaining sensitive information.  ...  the Gram-Schmidt process for each query.  ... 
fatcat:ynmzdc4usnet5o7iwyjdneu6bi

The Franz-Parisi Criterion and Computational Trade-offs in High Dimensional Statistics [article]

Afonso S. Bandeira, Ahmed El Alaoui, Samuel B. Hopkins, Tselil Schramm, Alexander S. Wein, Ilias Zadik
2022 arXiv   pre-print
We define a free-energy based criterion for hardness and formally connect it to the well-established notion of low-degree hardness for a broad class of statistical problems, namely all Gaussian additive  ...  and provide new low-degree lower bounds for sparse linear regression which seem difficult to prove directly.  ...  The authors thank Cris Moore and the Santa Fe Institute for hosting the 2018 "Santa Fe Workshop on Limits to Inference in Networks and Noisy Data," where the initial ideas in this paper were formulated  ... 
arXiv:2205.09727v1 fatcat:jauznr7cbbap5i66yhomkx24f4

Dagstuhl Reports, Volume 1, Issue 8, August 2011, Complete Issue [article]

2011
Data bootstrapping and consensus clustering was used to compute the robustness of branches and identify the most informative SNPs. 2.  ...  What are the bottlenecks of open science?  ...  Learning vector quantization is a robust prototypebased classification method which, together with the relevance learning strategy, assesses the relative contribution of spectral bands for efficient classification  ... 
doi:10.4230/dagrep.1.8 fatcat:5hegckr6vrbnfbssydweidcwzy

On incorporating inductive biases into deep neural networks [article]

Sameera Ramasinghe, University, The Australian National
2022
Despite the impressive performance, attention is being drawn towards enhancing the (relatively) weaker areas of deep models such as learning with limited resources, robustness, minimal overhead to realize  ...  Lars Petersson, for their continuous support of my Ph.D. research. Without their invaluable knowledge in computer vision, feedback, and guidance, I would not have been able to complete my studies.  ...  Robustness against information loss One critical requirement of a 3D object classification model is to be robust against information loss.  ... 
doi:10.25911/ph29-x543 fatcat:m5rkvjyy65hunkbp56qvgkpgqi
« Previous Showing results 1 — 15 out of 27 results