检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Liu, Yuhang Gong, Dong Gao, Erdun Zhang, Zhen Huang, Biwei Gong, Mingming Hengel, Anton van den Shi, Javen Qinfeng Australian Institute for Machine Learning The University of Adelaide Australia School of Computer Science and Engineering The University of New South Wales Australia Halicioglu Data Science Institute University of California San Diego United States School of Mathematics and Statistics The University of Melbourne Australia

The remarkable achievements of large language models (LLMs) have led many to conclude that they exhibit a form of intelligence. This is as opposed to explanations of their capabilities based on their ability to perform relatively simple manipulations of vast volumes of data. To illuminate the distinction between these explanations, we introduce a novel generative model that generates tokens on the basis of human-interpretable concepts represented as latent discrete variables. Under mild conditions—even when the mapping from the latent space to the observed space is non-invertible—we establish an identifiability result: the representations learned by LLMs through next-token prediction can be approximately modeled as the logarithm of the posterior probabilities of these latent discrete concepts, up to an invertible linear transformation. This theoretical finding not only provides evidence that LLMs capture underlying generative factors, but also strongly reinforces the linear representation hypothesis, which posits that LLMs learn linear representations of human-interpretable concepts. Empirically, we validate our theoretical results through evaluations on both simulation data and the Pythia, Llama, and DeepSeek model families. Copyright © 2025, The Authors. All rights reserved.

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

CUTE: Measuring LLMs' Understanding of Their Tokens

arXiv

引用

arXiv 2024年

作者： Edman, Lukas Schmid, Helmut Fraser, Alexander Center for Information and Language Processing LMU Munich Germany School of Computation Information and Technology TU Munich Germany Munich Center for Machine Learning Germany Munich Data Science Institute Germany

Large Language Models (LLMs) show remarkable performance on a wide variety of tasks. Most LLMs split text into multi-character tokens and process them as atomic units without direct access to individual characters. This raises the question: To what extent can LLMs learn orthographic information? To answer this, we propose a new benchmark, CUTE, which features a collection of tasks designed to test the orthographic knowledge of LLMs. We evaluate popular LLMs on CUTE, finding that most of them seem to know the spelling of their tokens, yet fail to use this information effectively to manipulate text, calling into question how much of this knowledge is generalizable. Copyright © 2024, The Authors. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

New Approach of Artificial Intelligence in Digital Forensic Investigation: A Literature Review 1

引用

International Conference on Advances in Communication Technology and Computer Engineering, ICACTCE 2023

作者： Verma, Raghav Garg, Sonia Kumar, Karan Gupta, Gaurav Salehi, Waleed Pareek, Piyush Kumar Kniežova, Jaroslava Yogananda School of Artificial Intelligence Computers and Data Science Himachal Pradesh Solan173229 India Mullana Ambala133207 India Department of Artificial Intelligence and Machine Learning Nitte Meenakshi Institute of Technology Bengaluru India Faculty of Management Comenius University in Bratislava 25 Odbojárov 10 Bratislava82005 Slovakia

ISBN: (数字)9783031371646

ISBN: (纸本)9783031371639

A Digital Forensic is a subfield of forensic science known as "digital forensic science" that focuses on the recovery and examination of data from digital devices that are connected to cybercrime. Computer forensics was the original meaning of the word "digital forensics." it has broadened to include any device that could store digital data. National policies on digital forensics didn't start to arise until the beginning of the twenty-first century. The process of locating, protecting, analyzing, and documenting digital evidence is known as digital forensics. When necessary, this is done so that evidence can be presented in court. AI in Digital forensics is one of the technologies that is developing the fastest, and it has had a significant impact on the methods and tools used to examine, monitor, and visualize crime scenes as well as develop effective strategies for dealing with impending threats and attacks on the internet and cyberspace. Modern technology has replaced human labor with machine-oriented designed work that has the capability and ability to minimize error and increase quality, reducing human effort and achieving maximum results with the fewest possible errors. It can influence the outcome and analyze the evidence in the field of digital forensics in a more effective manner to track the outcomes. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Artificial intelligence

来源：评论

学校读者我要写书评

暂无评论

Investigation of Red Fox Algorithm for Solving Task Assignment Problem in Heterogeneous Unmanned Aerial Vehicle Swarm

Investigation of Red Fox Algorithm for Solving Task Assignme...

引用

Artificial Intelligence For Internet of Things (AIIoT), International Conference on

作者： K. Manaswitha Titus Issac Salaja Silas J Sebastian Terance Division of Data Science and Cyber Security Karunya Institute of Technology and Sciences Coimbatore India Division of Computer Science and Engineering Karunya Institute of Technology and Sciences Coimbatore India Division of Artificial Intelligence and Machine Learning Karunya Institute of Technology and Sciences Coimbatore India

ISBN: (数字)9798350372120

ISBN: (纸本)9798350372137

In recent years, there have been significant advancements in Unmanned Aerial Vehicles (UAVs), leading to their integration into everyday life. UAVs are constrained by limited energy, communication, and localization. The various challenges in UAV research include localization, routing, navigation, and task assignment. Recently, UAV swarms have been employed across a wide range of applications. This study focuses on solving task assignment problems within UAV swarms, considering the unique multi-objective resource constraints of UAVs. The work explores the feasibility of adopting the Red Fox Algorithm to address the task assignment problem in UAV swarms. The investigation examines two primary input parameters: Task Energy Demand and distance. The evaluation includes energy utilization and shortest distance as performance metrics. Notably, the proposed approach outperforms the Particle Swarm Optimization Algorithm (PSO), resulting in a 17.47% lower energy utilization while maintaining a 17.83% reduction in the shortest distance travelled. By leveraging the Red Fox Algorithm, the work aims to enhance the efficiency and effectiveness of task allocation in UAV swarms, contributing to their broader adoption across various applications.

关键词： Location awareness Measurement Navigation Autonomous aerial vehicles Routing Resource management Internet of Things

来源：评论

学校读者我要写书评

暂无评论

On the Relation between Fourier and Walsh-Rademacher Spectra for Isotropic Random Fields

SSRN

引用

SSRN 2022年

作者： Kutsenko, Anton A. Danilov, Sergey D. Juricke, Stephan Oliver, Marcel Alfred Wegener Institute for Polar and Marine Research Bremerhaven Germany Jacobs University Bremen Germany Mathematical Institute for Machine Learning and Data Science KU Eichstätt Ingolstadt Germany

We discuss relations between Fourier coefficients corresponding to different bases having some scaling parameters. The focus will be on two popular basic systems: Walsh-Rademacher basis and trigonometric Fourier basis. We show that, generally, the statistical correlation between the Fourier coefficients corresponding to these systems is high enough for a qualitative analysis of the energy spectrum of wave fields in both bases, but, at the same time, the systems may give different results in, e.g., high frequency regimes and in over-dissipated models. The results are based on rigorous statistical and deterministic analysis. © 2022, The Authors. All rights reserved.

关键词： Fourier analysis

来源：评论

学校读者我要写书评

暂无评论

Intelligent Posture Detection System for Improved Ergonomics

Intelligent Posture Detection System for Improved Ergonomics

引用

2025 International Conference on Intelligent Control, Computing and Communications, IC3 2025

作者： Patil, Jay Joshi, Suvarna Mulla, Rahesha Yadav, Kushal Mit School of Computing Mit Adt University Department of Computer Science and Eng. Pune India University Department of Artificial Intelligence and Machine Learning Pune India Indian Institute of Technology Madras Department of Data Science India

ISBN: (纸本)9798331527495

This paper introduces a cost-effective smart posture detection system aimed at promoting ergonomic sitting habits and preventing musculoskeletal disorders. The system integrates ultrasonic sensors for backrest proximity measurement and load cells for seat weight distribution analysis. By analyzing these inputs, it identifies posture irregularities and provides real-time feedback for corrective action. Testing results demonstrate the system's reliability, achieving a detection accuracy of 99% across users with varying body types and postures. The proposed hybrid approach ensures simplicity, affordability, and scalability, making it a viable solution for homes, workplaces, and schools. This research contributes to the growing field of smart health monitoring systems, offering a practical alternative to traditional posture correction methods. © 2025 IEEE.

关键词： Ultrasonic testing

来源：评论

学校读者我要写书评

暂无评论

The Implicit Regularization of Dynamical Stability in Stochastic Gradient Descent

arXiv

引用

arXiv 2023年

作者： Wu, Lei Su, Weijie J. School of Mathematical Sciences Peking University Beijing China Center for Machine Learning Research Peking University Beijing China Wharton Statistics and Data Science Department University of Pennsylvania Philadelphia United States

In this paper, we study the implicit regularization of stochastic gradient descent (SGD) through the lens of dynamical stability (Wu et al., 2018). We start by revising existing stability analyses of SGD, showing how the Frobenius norm and trace of Hessian relate to different notions of stability. Notably, if a global minimum is linearly stable for SGD, then the trace of Hessian must be less than or equal to 2/η, where η denotes the learning rate. By contrast, for gradient descent (GD), the stability imposes a similar constraint but only on the largest eigenvalue of Hessian. We then turn to analyze the generalization properties of these stable minima, focusing specifically on two-layer ReLU networks and diagonal linear networks. Notably, we establish the equivalence between these metrics of sharpness and certain parameter norms for the two models, which allows us to show that the stable minima of SGD provably generalize well. By contrast, the stability-induced regularization of GD is provably too weak to ensure satisfactory generalization. This discrepancy provides an explanation of why SGD often generalizes better than GD. Note that the learning rate (LR) plays a pivotal role in the strength of stability-induced regularization. As the LR increases, the regularization effect becomes more pronounced, elucidating why SGD with a larger LR consistently demonstrates superior generalization capabilities. Additionally, numerical experiments are provided to support our theoretical findings. © 2023, CC BY-SA.

关键词： Stability

来源：评论

学校读者我要写书评

暂无评论

Sentiment Analysis Using machine learning Methods

Sentiment Analysis Using Machine Learning Methods

引用

2024 International Conference on Intelligent and Innovative Practices in Engineering and Management, IIPEM 2024

作者： Badholia, Abhishek Diwan, Tarun Dhar Narooka, Preeti Khatkale, Pravin B Vishnoi, Ankit Kaushik, Keshav Shri Shankaracharya Institute of Professional Management and Technology Department of Data Science Raipur India Atal Bihari Vajpayee University Bilaspur India Manipal University Artificial Intelligence and Machine Learning School of Computer and Engineering Jaipur India Sanjivani University Maharashtra Kopargaon India Graphic Era Deemed to Be University Department of Computer Science and Engineering Uttarakhand Dehradun India Amity University Punjab Amity School of Engineering and Technology Mohali India

ISBN: (纸本)9798350390049

machine learning (ML) will be utilized to evaluate sentiment analysis in this project. This study aims to do this. Sentiment analysis is a popular natural language processing area. This is a natural language processing subfield. This section identifies textual senses. Due to the growing availability of digital text across platforms, sentiment analysis is now vital for many applications. This is because sentiment analysis has become essential. This technology might be used for market research, political analysis, and consumer feedback. A few instances. This study analyzes several machine learning methods, to name a few. This category includes SVM, Naive Bayes, Random Forest, and RNNs. To clarify, several datasets are used to assess these methodologies. To get more information. Preprocessing and feature engineering are crucial to sentiment analysis model construction, as this paper shows. The research emphasises these two phases. In addition, it provides insights into how these approaches are used in constantly changing digital situations. © 2024 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Homophily and Incentive Effects in Use of Algorithms 44

Homophily and Incentive Effects in Use of Algorithms

引用

44th Annual Meeting of the Cognitive science Society: Cognitive Diversity, CogSci 2022

作者： Fogliato, Riccardo Fazelpour, Sina Gupta, Shantanu Lipton, Zachary Danks, David Department of Statistics and Data Science Carnegie Mellon University United States Department of Philosophy and Religion Khoury College of Computer Sciences Northeastern University United States Machine Learning Department Carnegie Mellon University United States Halicioğlu Data Science Institute Department of Philosophy University of California San Diego United States

As algorithmic tools increasingly aid experts in making consequential decisions, the need to understand the precise factors that mediate their influence has grown commensurately. In this paper, we present a crowdsourcing vignette study designed to assess the impacts of two plausible factors on AI-informed decision-making. First, we examine homophily-do people defer more to models that tend to agree with them?-by manipulating the agreement during training between participants and the algorithmic tool. Second, we considered incentives-how do people incorporate a (known) cost structure in the hybrid decision-making setting?-by varying rewards associated with true positives vs. true negatives. Surprisingly, we found limited influence of either homophily and no evidence of incentive effects, despite participants performing similarly to previous studies. Higher levels of agreement between the participant and the AI tool yielded more confident predictions, but only when outcome feedback was absent. These results highlight the complexity of characterizing human-algorithm interactions, and suggest that findings from social psychology may require re-examination when humans interact with algorithms. © 2022 The Author(s). This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY)

关键词： machine learning

来源：评论

学校读者我要写书评

暂无评论

A Closer Look at Benchmarking Self-Supervised Pre-training with Image Classification

arXiv

引用

arXiv 2024年

作者： Marks, Markus Knott, Manuel Kondapaneni, Neehar Cole, Elijah Defraeye, Thijs Perez-Cruz, Fernando Perona, Pietro California Institute of Technology United States ETH Zurich Institute for Machine Learning Department of Computer Science Switzerland Swiss Data Science Center ETH Zurich and EPFL Switzerland Empa Swiss Federal Laboratories for Materials Science and Technology Switzerland Altos Labs Switzerland

Self-supervised learning (SSL) is a machine learning approach where the data itself provides supervision, eliminating the need for external labels. The model is forced to learn about the data's inherent structure or context by solving a pretext task. With SSL, models can learn from abundant and cheap unlabeled data, significantly reducing the cost of training models where labels are expensive or inaccessible. In Computer Vision, SSL is widely used as pre-training followed by a downstream task, such as supervised transfer, few-shot learning on smaller labeled data sets, and/or unsupervised clustering. Unfortunately, it is infeasible to evaluate SSL methods on all possible downstream tasks and objectively measure the quality of the learned representation. Instead, SSL methods are evaluated using in-domain evaluation protocols, such as fine-tuning, linear probing, and k-nearest neighbors (kNN). However, it is not well understood how well these evaluation protocols estimate the representation quality of a pre-trained model for different downstream tasks under different conditions, such as dataset, metric, and model architecture. In this work, we study how classification-based evaluation protocols for SSL correlate and how well they predict downstream performance on different dataset types. Our study includes eleven common image datasets and 26 models that were pre-trained with different SSL methods or have different model backbones. We find that in-domain linear/kNN probing protocols are, on average, the best general predictors for out-of-domain performance. We further investigate the importance of batch normalization for the various protocols and evaluate how robust correlations are for different kinds of dataset domain shifts. In addition, we challenge assumptions about the relationship between discriminative and generative self-supervised methods, finding that most of their performance differences can be explained by changes to model backbones. Copyright © 2024, The

关键词： Nearest neighbor search

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：