检索结果-内蒙古大学图书馆

Privacy Amplification by Structured Subsampling for Deep Differentially Private Time Series Forecasting

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Schuchardt, Jan Dalirrooyfard, Mina Guzelkabaagac, Jed Schneider, Anderson Nevmyvaka, Yuriy Günnemann, Stephan Technical University of Munich Munich Data Science Institute Germany Machine Learning Research Morgan Stanley United States

Many forms of sensitive data, such as web traffic, mobility data, or hospital occupancy, are inherently sequential. The standard method for training machine learning models while ensuring privacy for units of sensitive information, such as individual hospital visits, is differentially private stochastic gradient descent (DP-SGD). However, we observe in this work that the formal guarantees of DP-SGD are incompatible with timeseries-specific tasks like forecasting, since they rely on the privacy amplification attained by training on small, unstructured batches sampled from an unstructured dataset. In contrast, batches for forecasting are generated by (1) sampling sequentially structured time series from a dataset, (2) sampling contiguous subsequences from these series, and (3) partitioning them into context and ground-truth forecast windows. We theoretically analyze the privacy amplification attained by this structured subsampling to enable the training of forecasting models with sound and tight event- and user-level privacy guarantees. Towards more private models, we additionally prove how data augmentation amplifies privacy in self-supervised training of sequence models. Our empirical evaluation demonstrates that amplification by structured subsampling enables the training of forecasting models with strong formal privacy guarantees. Copyright © 2025, The Authors. All rights reserved.

关键词： Differential privacy

TempXAI: Explainable AI for Time Series and data Streams 2024

学校读者我要写书评

暂无评论

TempXAI: Explainable AI for Time Series and Data Streams 202...

2024 Workshop on Explainable AI for Time Series and data Streams, TempXAI 2024

作者： Abdallah, Zahraa Fumagalli, Fabian Hammer, Barbara Hüllermeier, Eyke Jakobs, Matthias Müller, Emmanuel Muschalik, Maximilian Papapetrou, Panagiotis Saadallah, Amal Tzagkarakis, George University of Bristol United Kingdom Bielefeld University Germany LMU Munich Germany Lamarr Institute for Machine Learning and Artificial Intelligence Germany Research Center Trustworthy Data Science and Security Germany Stockholm University Sweden FORTH-ICS Greece

This document is the preface of the proceedings of the TempXAI Workshop Tutorial on Explainable AI for Time Series and data Streams, held on September 9th, 2024, in Vilnius, Lithuania. We received 5 submissions for peer-review, out of which we accepted 4 papers for this volume. In addition, we publish two extended abstracts of the tutorial that we give as a part of the workshop program. © 2024 CEUR-WS. All rights reserved.

关键词： data streams

Combining Evidence Across Filtrations

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Choe, Yo Joong Ramdas, Aaditya Data Science Institute University of Chicago United States Department of Statistics and Data Science Machine Learning Department Carnegie Mellon University United States

In sequential anytime-valid inference, any admissible procedure must be based on e-processes: generalizations of test martingales that quantify the accumulated evidence against a composite null hypothesis at any stopping time. This paper proposes a method for combining e-processes constructed in different filtrations but for the same null. Although e-processes in the same filtration can be combined effortlessly (by averaging), e-processes in different filtrations cannot because their validity in a coarser filtration does not translate to a finer filtration. This issue arises in sequential tests of randomness and independence, as well as in the evaluation of sequential forecasters. We establish that a class of functions called adjusters can lift arbitrary e-processes across filtrations. The result yields a generally applicable "adjust-then-combine" procedure, which we demonstrate on the problem of testing randomness in real-world financial data. Furthermore, we prove a characterization theorem for adjusters that formalizes a sense in which using adjusters is necessary. There are two major implications. First, if we have a powerful e-process in a coarsened filtration, then we readily have a powerful e-process in the original filtration. Second, when we coarsen the filtration to construct an e-process, there is a logarithmic cost to recovering validity in the original filtration.1 © 2024, CC BY.

关键词： Wiener filtering

Closed-form solutions for Bernoulli and compound Poisson branching processes in random environments

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Kutsenko, Anton A. University of Hamburg MIN Faculty Department of Mathematics Hamburg20146 Germany Mathematical Institute for Machine Learning and Data Science Katholische Universität Eichstätt Ingolstadt Germany

For branching processes, the generating functions for limit distributions of so-called ratios of probabilities of rare events satisfy the Schröder-type integral-functional equations. Excepting limited special cases, the corresponding equations can not be solved analytically. I found a large class of Poisson-type offspring distributions, for which the Schröder-type functional equations can be solved analytically. Moreover, for the asymptotics of limit distributions, the power and constant factor can be written explicitly. As a bonus, Bernoulli branching processes in random environments are treated. The beauty of this example is that the explicit formula for the generating function is unknown. Still, the closed-form expression for the power and constant factor in the asymptotic can be written with the help of some outstanding tricks. Also, the asymptotic expansion contains oscillatory terms absent in the Poisson case. It is shown that at the power 3 in the Bernoulli binomial kernels short-phase discrete oscillations turn into long-phase continuous ones. Copyright © 2024, The Authors. All rights reserved.

关键词： Poisson distribution

Stability Bounds for Smooth Optimal Transport Maps and their Statistical Implications

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Balakrishnan, Sivaraman Manole, Tudor Department of Statistics and Data Science Machine Learning Department Carnegie Mellon University United States Statistics and Data Science Center Massachusetts Institute of Technology United States

We study estimators of the optimal transport (OT) map between two probability distributions. We focus on plugin estimators derived from the OT map between estimates of the underlying distributions. We develop novel stability bounds for OT maps which generalize those in past work, and allow us to reduce the problem of optimally estimating the transport map to that of optimally estimating densities in the Wasserstein distance. In contrast, past work provided a partial connection between these problems and relied on regularity theory for the Monge-Ampère equation to bridge the gap, a step which required unnatural assumptions to obtain sharp guarantees. We also provide some new insights into the connections between stability bounds which arise in the analysis of plugin estimators and growth bounds for the semi-dual functional which arise in the analysis of Brenier potential-based estimators of the transport map. We illustrate the applicability of our new stability bounds by revisiting the smooth setting studied by Manole et al. [2024], analyzing two of their estimators under more general conditions. Critically, our bounds do not require smoothness or boundedness assumptions on the underlying measures. As an illustrative application, we develop and analyze a novel tuning parameter-free estimator for the OT map between two strongly log-concave distributions. Copyright © 2025, The Authors. All rights reserved.

关键词： Probability distributions

Impact of Immersive Technology on Physiological Parameters in Psychiatric Disorders

学校读者我要写书评

暂无评论

Impact of Immersive Technology on Physiological Parameters i...

International Conference on System Modeling & Advancement in Research Trends (SMART)

作者： Anjali Kataria Gurjinder Singh Jasminder Sandhu Chitkara Institute of Engineering and Technology Punjab India Department of Machine Learning and Data Science IILM University Noida

ISBN: (数字)9798350380583

ISBN: (纸本)9798350380590

The brain is an essential component that regulates the general functioning of the body. The brain consists of millions of neurons that govern human behavior in response to sensory stimuli. To understand the mental functioning of the brain, it is important to assess its neural signals. Presently, the exponential expansion of prescription medications and substances, including alcohol and drugs, is exerting a significant influence on human psychology. The rising prevalence of different health disorders can be attributed to the increasing number of cases. The sensorimotor cortex, which in humans is in charge of coordinating both physical and cognitive activities, can be used to visualize brain signals. The present research describes an empirical investigation that employs Virtual Reality therapy to examine the affective condition of individuals suffering from drug addiction. The study utilized the Emotiv Insight 5-channel EEG headset to collect EEG data from 50 participants, and the Emotiv Pro software was employed for data analysis. The study findings indicate that Virtual Reality treatment has a substantial positive impact on the stress levels and concentration of persons suffering from drug addiction.

关键词： Drugs Wavelet transforms Addiction Neurons Psychology Medical treatment Virtual reality Electroencephalography Software Physiology

The Mathematics of Dots and Pixels: On the Theoretical Foundations of Image Halftoning

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Krahmer, Felix Veselovska, Anna Technical University of Munich Department of Mathematics and Munich Data Science Institute Munich Center for Machine Learning Germany

The evolution of image halftoning, from its analog roots to contemporary digital methodologies, encapsulates a fascinating journey marked by technological advancements and creative innovations. Yet the theoretical understanding of halftoning is much more recent. In this article, we explore various approaches towards shedding light on the design of halftoning approaches and why they work. We discuss both halftoning in a continuous domain and on a pixel grid. We start by reviewing the mathematical foundation of the so-called electrostatic halftoning method, which departed from the heuristic of considering the back dots of the halftoned image as charged particles attracted by the grey values of the image in combination with mutual repulsion. Such an attraction-repulsion model can be mathematically represented via an energy functional in a reproducing kernel Hilbert space allowing for a rigorous analysis of the resulting optimization problem as well as a convergence analysis in a suitable topology. A second class of methods that we discuss in detail is the class of error diffusion schemes, arguably among the most popular halftoning techniques due to their ability to work directly on a pixel grid and their ease of application. The main idea of these schemes is to choose the locations of the black pixels via a recurrence relation designed to agree with the image in terms of the local averages. We discuss some recent mathematical understanding of these methods that is based on a connection to ΣΔ quantizers, a popular class of algorithms for analog-to-digital *** Codes 94A08, 62H35 © 2024, CC BY-NC-SA.

关键词： Pixels

A Comprehensive Review of Supervised learning Algorithms in Healthcare Applications 2

学校读者我要写书评

暂无评论

A Comprehensive Review of Supervised Learning Algorithms in ...

2nd DMIHER International Conference on Artificial Intelligence in Healthcare, Education and Industry, IDICAIEI 2024

作者： Barhate, Aditya Tale, Abhay Jikar, Nayan Verma, Prateek Kumar, Praveen Yesankar, Prajyot Datta Meghe Institute of Higher Education and Research Faculty of Engineering and Technology Department of Artificial Intelligence and Machine Learning Maharashtra Wardha India Datta Meghe Institute of Higher Education and Research Faculty of Engineering and Technology Department of Artificial Intelligence and Data Science Maharashtra Wardha India

ISBN: (纸本)9798331528713

Supervised learning has revolutionized the concept of personalization in treatment with the development of Precision Medicine. This review aims to provide a systematic analysis of the utilization of supervised learning in healthcare and its functions in diagnostics and prognoses, recommendations for treatment, and patient surveillance. Classification algorithms such as Logistic Regression, Support Vector machines, and Decision Trees. Regression techniques such as Linear Regression, Polynomial Regression, and Neural Networks. Ensemble methods such as Bagging, Boosting, and Stacking. Deep learning techniques such as Convolutional Neural Networks, Recurrent Neural Networks, and Long Short-Term Memory are discussed for their efficiency and applicability in various healthcare applications. There are strengths highlighted by the review including enhanced diagnostic capabilities and patient-targeted therapy alongside weaknesses such as quality of data, interpretability as well as compatibility with existing healthcare systems. This review gives a more generalized approach to how supervised learning algorithms help to enhance healthcare results and identifies the likely topic that requires more research in the future. © 2024 IEEE.

关键词： Self-supervised learning

DETERMINISTIC AND STOCHASTIC SURROGATE MODELS FOR A SLOWLY DRIVEN FAST OSCILLATOR

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Oliver, Marcel Kenfack, Marc Aurele Tiofack Mathematical Institute for Machine Learning and Data Science KU Eichstätt–Ingolstadt Ingolstadt85049 Germany Constructor University Bremen28759 Germany

It has long been known that the excitation of fast motion in certain two-scale dynamical systems is linked to the singularity structure in complex time of the slow variables. We demonstrate, in the context of a fast harmonic oscillator forced by one component of the Lorenz 1963 model, that this principle can be used to construct time-discrete surrogate models by numerically extracting approximate locations and residues of complex poles via Adaptive Antoulas–Andersen (AAA) rational interpolation and feeding this information into the known "connection formula" to compute the resulting fast amplitude. Despite small but nonnegligible local errors, the surrogate model maintains excellent accuracy over very long times. In addition, we observe that the long-time behavior of fast energy offers a continuous-time analog of Gottwald and Melbourne’s 2004 "0–1 test for chaos" – the asymptotic growth rate of the energy in the oscillator can discern whether or not the forcing function is *** Codes 34M40, 41A20, 34E13 Copyright © 2023, The Authors. All rights reserved.

关键词： Stochastic models