检索结果-内蒙古大学图书馆

37th Annual Conference on Learning Theory, COLT 2024

作者： Duchi, John Haque, Saminul Departments of Statistics and Electrical Engineering Stanford University United States Department of Computer Science Stanford University United States

We present an information-theoretic lower bound for the problem of parameter estimation with time-uniform coverage guarantees. Via a new a reduction to sequential testing, we obtain stronger lower bounds that capture the hardness of the time-uniform setting. In the case of location model estimation, logistic regression, and exponential family models, our Ω(√n-1 log log n) lower bound is sharp to within constant factors in typical settings. © 2024 J. Duchi & S. Haque.

关键词： Logistic regression

来源：评论

学校读者我要写书评

暂无评论

Is Temperature Sample Efficient for Softmax Gaussian Mixture of Experts? 41

Is Temperature Sample Efficient for Softmax Gaussian Mixture...

引用

41st International Conference on Machine Learning, ICML 2024

作者： Nguyen, Huy Akbarian, Pedram Ho, Nhat Department of Statistics and Data Sciences United States Department of Electrical and Computer Engineering The University of Texas at Austin United States

Dense-to-sparse gating mixture of experts (MoE) has recently become an effective alternative to a well-known sparse MoE. Rather than fixing the number of activated experts as in the latter model, which could limit the investigation of potential experts, the former model utilizes the temperature to control the softmax weight distribution and the sparsity of the MoE during training in order to stabilize the expert specialization. Nevertheless, while there are previous attempts to theoretically comprehend the sparse MoE, a comprehensive analysis of the dense-to-sparse gating MoE has remained elusive. Therefore, we aim to explore the impacts of the dense-to-sparse gate on the maximum likelihood estimation under the Gaussian MoE in this paper. We demonstrate that due to interactions between the temperature and other model parameters via some partial differential equations, the convergence rates of parameter estimations are slower than any polynomial rates, and could be as slow as O(1/log(n)), where n denotes the sample size. To address this issue, we propose using a novel activation dense-to-sparse gate, which routes the output of a linear layer to an activation function before delivering them to the softmax function. By imposing linearly independence conditions on the activation function and its derivatives, we show that the parameter estimation rates are significantly improved to polynomial rates. Finally, we conduct a simulation study to empirically validate our theoretical results. Copyright 2024 by the author(s)

关键词： Maximum likelihood estimation

来源：评论

学校读者我要写书评

暂无评论

Universally Instance-Optimal Mechanisms for Private Statistical Estimation 37

Universally Instance-Optimal Mechanisms for Private Statisti...

引用

37th Annual Conference on Learning Theory, COLT 2024

作者： Asi, Hilal Duchi, John C. Haque, Saminul Li, Zewei Ruan, Feng Apple United States Departments of Statistics and Electrical Engineering Stanford University United States Department of Computer Science Stanford University United States Department of Statistics and Data Science Northwestern University United States

We consider the problem of instance-optimal statistical estimation under the constraint of differential privacy where mechanisms must adapt to the difficulty of the input dataset. We prove a new instance specific lower bound using a new divergence and show it characterizes the local minimax optimal rates for private statistical estimation. We propose two new mechanisms that are universally instance-optimal for general estimation problems up to logarithmic factors. Our first mechanism, the total variation mechanism, builds on the exponential mechanism with stable approximations of the total variation distance, and is universally instance-optimal in the high privacy regime Ε ≤ 1/√n. Our second mechanism, the T-mechanism, is based on the T-estimator framework (Birgé, 2006) using the clipped log likelihood ratio as a stable test: it attains instance-optimal rates for any Ε ≤ 1 up to logarithmic factors. Finally, we study the implications of our results to robust statistical estimation, and show that our algorithms are universally optimal for this problem, characterizing the optimal minimax rates for robust statistical estimation. © 2024 H. Asi, J.C. Duchi, S. Haque, Z. Li & F. Ruan.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Resampling methods for private statistical inference

arXiv

引用

arXiv 2024年

作者： Chadha, Karan Duchi, John C. Kuditipudi, Rohith Departments of Electrical Engineering Departments of Statistics Departments of Computer Science Stanford University United States

We propose two private variants of the non-parametric bootstrap for privately computing confidence sets. Each privately computes the median of results of multiple "little" bootstraps, yielding asymptotic bounds on the coverage error of the resulting confidence sets. For a fixed differential privacy parameter Ε, our methods enjoy the same error rates as the non-private bootstrap to within logarithmic factors in the sample size n. We empirically validate the performance of our methods for mean estimation, median estimation, and logistic regression, and our methods achieve similar coverage accuracy to existing methods (and non-private baselines) while providing notably shorter (about 10×) confidence intervals than previous approaches. Copyright © 2024, The Authors. All rights reserved.

关键词： Machine learning

来源：评论

学校读者我要写书评

暂无评论

Accelerated gradient methods for nonconvex optimization: Escape trajectories from strict saddle points and convergence to local minima

arXiv

引用

arXiv 2023年

作者： Dixit, Rishabh Gürbüzbalaban, Mert Bajwa, Waheed U. Department of Electrical and Computer Engineering Departments of Electrical & Computer Engineering Management Science and Information Systems and Statistics United States Departments of Electrical & Computer Engineering and Statistics

This paper considers the problem of understanding the behavior of a general class of accelerated gradient methods on smooth nonconvex functions. Motivated by some recent works that have proposed effective algorithms, based on Polyak’s heavy ball method and the Nesterov accelerated gradient method, to achieve convergence to a local minimum of nonconvex functions, this work proposes a broad class of Nesterov-type accelerated methods and puts forth a rigorous study of these methods encompassing the escape from saddle-points and convergence to local minima through a both asymptotic and a nonasymptotic analysis. In the asymptotic regime, this paper answers an open question of whether Nesterov’s accelerated gradient method (NAG) with variable momentum parameter avoids strict saddle points almost surely. This work also develops two metrics of asymptotic rate of convergence and divergence, and evaluates these two metrics for several popular standard accelerated methods such as the NAG, and Nesterov’s accelerated gradient with constant momentum (NCM) near strict saddle points. In the local regime, this work provides an analysis that leads to the "linear" exit time estimates from strict saddle neighborhoods for trajectories of these accelerated methods as well the necessary conditions for the existence of such trajectories. Finally, this work studies a sub-class of accelerated methods that can converge in convex neighborhoods of nonconvex functions with a near optimal rate to a local minima and at the same time this sub-class offers superior saddle-escape behavior compared to that of NAG. Copyright © 2023, The Authors. All rights reserved.

关键词： Gradient methods

来源：评论

学校读者我要写书评

暂无评论

A Phylogenetic Approach to Genomic Language Modeling 29th

A Phylogenetic Approach to Genomic Language Modeling

引用

29th International Conference on Research in Computational Molecular Biology, RECOMB 2025

作者： Albors, Carlos Li, Jianan Canal Benegas, Gonzalo Ye, Chengzhong Song, Yun S. Department of Electrical Engineering and Computer Sciences University of California BerkeleyCA94720 United States Department of Statistics University of California BerkeleyCA94720 United States

ISBN: (纸本)9783031902512

Genomic language models (gLMs) have shown mostly modest success in identifying evolutionarily constrained elements in mammalian genomes. To address this issue, we introduce a novel framework for training gLMs that explicitly models nucleotide evolution on phylogenetic trees using multispecies whole-genome alignments. Our approach integrates an alignment into the loss function during training but does not require it for making predictions, thereby enhancing the model’s applicability. We applied this framework to train PhyloGPN, a model that excels at predicting functionally disruptive variants from a single sequence alone and demonstrates strong transfer learning capabilities. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Transfer learning

来源：评论

学校读者我要写书评

暂无评论

Low-Rank Gap Filling and Downscaling for SMAP Soil Moisture Datasets

Low-Rank Gap Filling and Downscaling for SMAP Soil Moisture ...

引用

作者： Beale, Kevin Bras, Rafael L. Romberg, Justin Department of Electrical and Computer Engineering Georgia Institute of Technology AtlantaGA United States Departments of Civil and Environmental Engineering and Earth and Atmospheric Sciences Georgia Institute of Technology AtlantaGA United States

Soil moisture is the linchpin of the surface hydrologic cycle, controlling the partitioning of water and energy fluxes at the surface. Without it, vegetation, and hence life on the solid Earth as we know it, would not exist. Understanding ecohydrology is understanding the availability of soil moisture to vegetation. Until recently, measuring soil moisture was difficult, expensive, intrusive, and local. NASA's Soil Moisture Active Passive (SMAP) mission changed that by providing global estimates at reasonable frequencies. Ecohydrology and many other hydrologic applications are best when high spatiotemporal resolution soil moisture datasets are available. The SMAP and SMAP-Sentinel soil moisture products currently possess contrasting spatial and temporal resolutions, but their coincident nature presents an opportunity to learn how to enhance the spatial resolution of SMAP retrievals to obtain a global, high spatiotemporal resolution dataset. However, a challenge in learning from SMAP-Sentinel data is the presence of missing pixels. In this work, we propose a low-rank approach to both gap-fill SMAP-Sentinel and downscale SMAP and evaluate its performance globally on both held-out SMAP-Sentinel data and measurements from SMAPVEX validation datasets. The proposed method outperformed baselines globally on SMAP-Sentinel data but had mixed performance against retrievals from airborne measurements. A procedure for filling in missing pixels in SMAP-Sentinel measurements using the low-rank models was found to outperform alternative interpolation methods. Overall, the results show that the proposed method can recover missing pixels in soil moisture measurements and can be used to compute estimates of high-resolution SMAP-Sentinel retrievals from low-resolution SMAP data. © 2025 The Author(s). Ecohydrology published by John Wiley & Sons Ltd.

关键词： Orthogonal functions

来源：评论

学校读者我要写书评

暂无评论

IT-Security Risk Based Approach for Secure Operation of Distributed Data Platforms in Supply Chains

IT-Security Risk Based Approach for Secure Operation of Dist...

引用

2024 IEEE International Conference on Industrial engineering and engineering Management, IEEM 2024

作者： Voß, Marvin Kallisch, Jonas Runge, Maxim Theus, Tobias Niemann, Karl-Heinz Wunck, Christoph Faculty I - Electrical Engineering and Information Technology University of Applied Sciences and Arts Hannover Germany Department of Engineering Section Electrical Engineering and Computer Science University of Applied Sciences Leer Emden Germany Faculty of Mathematics Informatics and Statistics Ludwig Maximilian University Munich Germany

ISBN: (纸本)9798350386097

This paper presents the development of a secure data platform designed to enhance operational efficiency and to facilitate cross-company collaboration within the manufacturing supply chain. The platform is designed to ease secure information exchange while maintaining the sovereignty of each participant, leveraging standardized security features and advanced technologies such as federated learning. Security measures are based on a comprehensive risk analysis, which methodically identifies, assesses, and mitigates potential threats. Moreover, federated learning models that are not transparent are utilized. This allows data owners to maintain control over sensitive information within the consortium. The described approach markedly enhances both security and operational collaboration across the industry, representing a crucial stride towards more integrated and efficient supply chain management in manufacturing. © 2024 IEEE.

关键词： Risk analysis

来源：评论

学校读者我要写书评

暂无评论

Supervised learning applied to electrocardiogram statistical features for the detection of premature ventricular contraction

引用

Research on Biomedical engineering 2025年第1期41卷 1-12页

作者： Issa, Khouloud Rammal, Abbas Assaf, Rabih Ghandour, Ahmad Electrical and Communication Engineering Department College of Engineering Phoenicia University Zahrani South Lebanon Statistics and Informatics Department Faculty of Science Lebanese University Beirut Lebanon Mathematics and Computer Sciences Department Faculty of Arts and Sciences Lebanese American University Beirut Lebanon Department of Mathematics Faculty of Arts and Sciences Holy Spirit University of Kaslik Jounieh Lebanon Biomedical Engineering Department Faculty of Engineering Islamic University of Lebanon Wardenieh30014 Lebanon

Purpose: The development of an automated premature ventricular contraction (PVC) detection system has significant implications for early intervention and treatment decisions. This study aims to develop a novel approach using various supervised machine learning (ML) methods for detecting PVCs in electrocardiogram (ECG) recordings. Methodology: To achieve this, we extracted ten distinct statistical and temporal features from 33 long-term ECG recordings from the benchmark MIT-BIH arrhythmia database (MIT-BIH-AD), capturing significant characteristics from the signals. We then investigated the effectiveness of traditional ML algorithms in identifying PVCs based on these features, namely support vector machine (SVM), decision tree (DT), naïve Bayes (NB), k-nearest neighbor (KNN), linear discriminant analysis (LDA), and artificial neural network (ANN). Results: Among these classifiers, the SVM, KNN, ANN, and DT classifiers demonstrated exceptional discriminatory power and classification performance, yielding near-perfect area under the receiver operating characteristic curve (AUC-ROC) values of 97.3%, 95.4%, 95%, and 94.1%, respectively. On the other hand, the SVM classifier emerged as the most accurate, achieving an overall accuracy of 97.39%, indicative of making correct predictions. Conclusion: These findings underscore the generalization of the proposed approach, especially since we worked on an extended dataset encompassing different types of heartbeats to discriminate PVCs from other heartbeat types, not only normal ones, and providing closer alignment with real-world scenarios. Subsequently, the robustness and accuracy of these models highlight their suitability for clinical applications and their potential for efficient implementation and deployment in the diagnosis of cardiovascular diseases (CVDs), as they demand minimal computational resources and time. © The Author(s), under exclusive licence to The Brazilian Society of Biomedical engineering 2025.

关键词： Decision trees

来源：评论

学校读者我要写书评

暂无评论

Advances in random topology

引用

Journal of Applied and Computational Topology 2024年第6期8卷 1445-1448页

作者： Bobrowski, Omer Yogeshwaran, D. School of Mathematical Sciences Queen Mary University of London Viterbi Faculty of Electrical and Computer Engineering Technion – Israel Institute of Technology Theoretical Statistics and Mathematics Unit Indian Statistical Institute

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：