检索结果-内蒙古大学图书馆

Calibrating a Deep Neural Network with Its Predecessors

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Tao, Linwei Dong, Minjing Liu, Daochang Sun, Changming Xu, Chang School of Computer Science Faculty of Engineering University of Sydney Australia CSIRO's Data61 Australia

Confidence calibration - the process to calibrate the output probability distribution of neural networks - is essential for safety-critical applications of such networks. Recent works verify the link between mis-calibration and overfitting. However, early stopping, as a well-known technique to mitigate overfitting, fails to calibrate networks. In this work, we study the limitions of early stopping and comprehensively analyze the overfitting problem of a network considering each individual block. We then propose a novel regularization method, predecessor combination search (PCS), to improve calibration by searching a combination of best-fitting block predecessors, where block predecessors are the corresponding network blocks with weight parameters from earlier training stages. PCS achieves the state-of-the-art calibration performance on multiple datasets and architectures. In addition, PCS improves model robustness under dataset distribution shift. Supplementary material and code are available at https://***/Linwei94/PCS. © 2023, CC BY.

关键词： Calibration

Comparing Time-Series Analysis Approaches Utilized in Research Papers to Forecast COVID-19 Cases in Africa: A Literature Review

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Ebadi, Ali Sahafizadeh, Ebrahim Department of Computer Engineering Faculty of Intelligent Systems Engineering and Data Science Persian Gulf University Bushehr Iran

This literature review aimed to compare various time-series analysis approaches utilized in forecasting COVID-19 cases in Africa. The study involved a methodical search for English-language research papers published between January 2020 and July 2023, focusing specifically on papers that utilized time-series analysis approaches on COVID-19 datasets in Africa. A variety of databases including PubMed, Google Scholar, Scopus, and Web of science were utilized for this process. The research papers underwent an evaluation process to extract relevant information regarding the implementation and performance of the time-series analysis models. The study highlighted the different methodologies employed, evaluating their effectiveness and limitations in forecasting the spread of the virus. The result of this review could contribute deeper insights into the field, and future research should consider these insights to improve time series analysis models and explore the integration of different approaches for enhanced public health decision-making. © 2023, CC BY-NC-ND.

关键词： Coronavirus

Implementation of Face Patterns and Smile Recognition for Intelligent Class Attendance Systems

学校读者我要写书评

暂无评论

Implementation of Face Patterns and Smile Recognition for In...

International Conference of computer and Informatics Engineering (IC2IE)

作者： Miftakhurrokhmat Teddy Suparyanto Gregorius Natanael Elwirehardja Shofiyati Nur Karimah Bens Pardamean Bioinformatics and Data Science Research Center Bina Nusantara University Jakarta Indonesia Computer Science Department School of Computer Science Bina Nusantara University Jakarta Indonesia Department of Informatics Faculty of Computer Science Universitas Amikom Yogyakarta Yogyakarta Indonesia Computer Science Department BINUS Graduate Program - Master of Computer Science Bina Nusantara University Jakarta Indonesia

Students “attendance in class is one important success parameter in face-to-face learning processes. Conventional attendance systems, such as paper-based attendance sheets or identity card systems, require a long time in the manual recapitulation process. Without additional verifications, even computer vision-based methods are prone to fraudulent practices by the students instead of gaining their excitement and attention in a class. To stimulate students” attention in a class, this work designs an intelligent class attendance system, in which facial pattern and smile recognition are implemented by using the latter as an additional task-based verification to reduce the risks of fake attendance. For the face recognition module, this pilot study used FaceNet as a feature extractor combined with SVM for classification, whereas the Haar cascade algorithm is used for recognizing smiles. This face recognition pipeline was implemented as a service installed on minicomputers or Internet of Things (IoT) devices in each classroom and connected to an IP camera. Every recorded attendance will be sent as a notification to a mobile application for students that requires their active participation to confirm it with a smiling self-photo. The proposed pipeline obtained 92.86% accuracy on the test data, and 66.67% accuracy when evaluated in a real-life simulation setting through the implemented system. The lower accuracy in the simulation indicated that further improvements are indispensable, especially since the model obtained 28.57% False Negative Rate. Future studies will need to acquire more data and experiment with more efficient methods of attendance verification.

关键词：

Outlier Robust Multivariate Polynomial Regression 32

学校读者我要写书评

暂无评论

Outlier Robust Multivariate Polynomial Regression

32nd Annual European Symposium on Algorithms, ESA 2024

作者： Arora, Vipul Bhattacharyya, Arnab Boban, Mathews Guruswami, Venkatesan Kelman, Esty School of Computing National University of Singapore Singapore Department of EECS Department of Mathematics University of California BerkeleyCA United States CSAIL Massachusetts Institute of Technology CambridgeMA United States Department of Computer Science Faculty of Computing & Data Sciences Boston University MA United States

ISBN: (纸本)9783959773386

We study the problem of robust multivariate polynomial regression: let p: Rn → R be an unknown n-variate polynomial of degree at most d in each variable. We are given as input a set of random samples (xi, yi) ∈ [−1, 1]n × R that are noisy versions of (xi, p(xi)). More precisely, each xi is sampled independently from some distribution χ on [−1, 1]n, and for each i independently, yi is arbitrary (i.e., an outlier) with probability at most ρ n log d), where the hidden constant depends on n, if χ is the n-dimensional Chebyshev distribution. The sample complexity is On(d2n log d), if the samples are drawn from the uniform distribution instead. The approximation error is guaranteed to be at most O(σ), and the run-time depends on log(1/σ). In the setting where each xi and yi are known up to N bits of precision, the run-time's dependence on N is linear. We also show that our sample complexities are optimal in terms of dn. Furthermore, we show that it is possible to have the run-time be independent of 1/σ, at the cost of a higher sample complexity. © Vipul Arora, Arnab Bhattacharyya, Mathews Boban, Venkatesan Guruswami, and Esty Kelman;licensed under Creative Commons License CC-BY 4.0.

关键词： Logistic regression

AI-Powered Real-Time Obstacle Detection for UAVs using Blockchain

学校读者我要写书评

暂无评论

AI-Powered Real-Time Obstacle Detection for UAVs using Block...

International Conference on Parallel Distributed and Grid Computing (PDGC)

作者： Aparna Kumari Patel Ansh Alkeshbhai Shah Akshat Nimesh Safaa Najah Saud Al-Humairi Prasun Kumar Sunil Kumar Prakriti Lohumi Department of Computer Science and Engineering Institute of Technology Nirma University Ahmedabad India Department of Computer Science and Engineering Faculty of Engineering Science and Technology Adani University Ahmedabad India Faculty of Information Sciences and Engineering Management and Science University Malaysia Department of Biotechnology SSET Sharda University Greater Noida U.P. India Department of AI and Data Science Guru Jambheshwar University of Science & Technology Hisar Haryana India Department of Artificial Intelligence and Data Science Bhagwan Parshuram Institute of Technology New Delhi Delhi India

ISBN: (数字)9798331521349

ISBN: (纸本)9798331521356

AI integration with UAVs has brought significant advancements in aerial control and operational safety, particularly in real-time obstacle detection—an essential aspect for navigating unknown environments. This work introduces an innovative AI-based solution for obstacle detection in UAVs, leveraging deep learning techniques to enhance precision and environmental awareness. The system’s architecture involves multiple layers, where UAVs first capture high-resolution images that undergo a processing pipeline including data pre-processing, augmentation, and labeling. A key element of this process is the use of Convolutional Neural Networks (CNNs) to train models capable of identifying obstacles across various terrains. To ensure the integrity and security of the data, especially in complex multi-UAV systems, blockchain technology is integrated. Utilizing Distributed Hash Tables (DHTs) and the Interplanetary File System (IPFS), this decentralized system creates a content-addressable database to store and authenticate unalterable records. Experimental analysis demonstrates that this system offers high accuracy in real-time obstacle detection, minimizing false positives and improving UAV safety.

关键词： Deep learning Autonomous aerial vehicles Real-time systems Blockchains Safety Reliability InterPlanetary File System Convolutional neural networks Artificial intelligence Drones

Private Language Models via Truncated Laplacian Mechanism

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Huang, Tianhao Yang, Tao Habernal, Ivan Hu, Lijie Wang, Di Lab China King Abdullah University of Science and Technology Saudi Arabia Nankai University China Research Center Trustworthy Data Science and Security The University Alliance Ruhr Faculty of Computer Science Ruhr University Bochum Germany

Deep learning models for NLP tasks are prone to variants of privacy attacks. To prevent privacy leakage, researchers have investigated word-level perturbations, relying on the formal guarantees of differential privacy (DP) in the embedding space. However, many existing approaches either achieve unsatisfactory performance in the high privacy regime when using the Laplacian or Gaussian mechanism, or resort to weaker relaxations of DP that are inferior to the canonical DP in terms of privacy strength. This raises the question of whether a new method for private word embedding can be designed to overcome these limitations. In this paper, we propose a novel private embedding method called the high dimensional truncated Laplacian mechanism. Specifically, we introduce a non-trivial extension of the truncated Laplacian mechanism, which was previously only investigated in one-dimensional space cases. Theoretically, we show that our method has a lower variance compared to the previous private word embedding methods. To further validate its effectiveness, we conduct comprehensive experiments on private embedding and downstream tasks using three datasets. Remarkably, even in the high privacy regime, our approach only incurs a slight decrease in utility compared to the non-private scenario. © 2024, CC BY.

关键词： Differential privacy

NCC: Neural concept compression for multilingual document recommendation[Formula presented]

学校读者我要写书评

暂无评论

NCC: Neural concept compression for multilingual document re...

作者： Tashu, Tsegaye Misikir Lenz, Marc Horváth, Tomáš ELTE – Eötvös Loránd University Faculty of Informatics Institute of Academia-Industry Cooperation Department of Data Science and Engineering Pázmány Péter sétány 1/C 1117 Budapest Hungary Pavol Jozef Šafárik University in Košice Faculty of Science Institute of Computer Science Jesenná 5 Košice040 01 Slovakia Department of Artificial Intelligence Bernoulli Institute of Mathematics Computer Science and Artificial Intelligence University of Groningen Groningen Netherlands Wollo University Kombolcha Institute of Technology College of Informatics Kombocha Ethiopia

In this work, we propose a novel method for generating inter-lingual document representations using neural network concept compression. The presented approach is intended to improve the quality of content-based multilingual document recommendation and information retrieval systems by creating a language-independent representation. The main idea is to use mappings to align monolingual representation spaces, using concept compression, to create inter-lingual representations. The proposed approach outperforms traditional cross-lingual retrieval and recommendations methods in experiments conducted on JRC-Acquis and EU bookshop multilingual corpora. Our dataset and code are publicly available at https://***/Tsegaye-misikir/NCC. © 2023

关键词： Information retrieval systems

An Evaluation and Implementation of Pretty Good Privacy (PGP) in Wireless Network Security 3

学校读者我要写书评

暂无评论

An Evaluation and Implementation of Pretty Good Privacy (PGP...

3rd International Conference on Smart Generation Computing, Communication and Networking, SMART GENCON 2023

作者： Hans, Arvind Vigenesh, M. Yuvaraj, S. Bhagwat, Suvarna R. Zaki, Hamza Yamang, Chiging Usha Martin University Faculty of Business Mangagement & Commerce Ranchi India Karpagam Academy of Higher Education Department of Computer Science Engineering Coimbatore India Prince Shri Venkateshwara Padmavathy Engineering College Department of Electronics and Communication Engineering Chennai India Vishwakarma Institute of Information Technology Department of Artificial Intelligence & Data Science Pune India Mangalayatan University Department of Computer Science Aligarh India Himalayan University Department of Bussiness & Management Itanagar India

ISBN: (纸本)9798350319125

Top privateers (PGP) is an encryption protocol that shields statistics saved, exchanged, or transmitted over wireless networks. PGP provides a green manner to comfy WiFi networks and communications using symmetric and publickey cryptography. This paper evaluates and examines the utility of PGP in wireless network protection. The paper critiques the layout and implementation of PGP on specific varieties of Wi-Fi networks, which include the partnerships to be had for implementing PGP in pre-present and destiny wireless infrastructures. Similarly to the numerous partnerships, the paper explores PGP's unique protocols and technologies and how they can be integrated inside Wi-Fi networks based on their extraordinary capabilities and functions. The paper additionally discusses the commercial advertising of PGP, its effect on wireless community protection, and the packages of PGP, in addition to a number of the demanding situations and threats related to its use. Moreover, the paper highlights the advantages of implementing PGP in wireless community safety. It gives a diffusion of trying out and implementation strategies that can be used to ensure a successful implementation. The paper concludes with a dialogue about the possible destiny directions of PGP and how it could be used to ease and protect wireless networks. © 2023 IEEE.

关键词： destiny infrastructures networks packages wireless networks

Avoiding exp(Rmax) scaling in RLHF through Preference-based Exploration

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Chen, Mingyu Chen, Yiding Sun, Wen Zhang, Xuezhou Department of Electrical & Computer Engineering Boston University United States Department of Computer Science Cornell University United States Faculty of Computing & Data Sciences Boston University United States

Reinforcement Learning from Human Feedback (RLHF) has emerged as a pivotal technique for large language model (LLM) alignment. This paper studies the setting of online RLHF and focus on improving sample efficiency. All existing algorithms in online RLHF, whether doing passive exploration or active exploration, suffer from a sample complexity that scales exponentially with the scale of the reward function. This fundamental limitation hinders their effectiveness in scenarios with heavily skewed preferences, e.g. questions with a unique correct solution. To address this, we introduce Self-Exploring Preference-Incentive Online Preference Optimization (SE-POPO), an online RLHF algorithm that for the first time achieves a sample complexity that scales polynomially with the reward scale, answering an open problem raised by Xie et al. (2024). Theoretically, we demonstrate that the sample complexity of SE-POPO dominates that of existing exploration algorithms. Empirically, our systematic evaluation confirms that SE-POPO is more sample-efficient than both exploratory and non-exploratory baselines, in two primary application scenarios of RLHF as well as on public benchmarks, marking a significant step forward in RLHF algorithm design. The code is available at https://***/MYC000801/SE-POPO. © 2025, CC BY.

关键词： Reinforcement learning