检索结果-内蒙古大学图书馆

arXiv 2022年

作者： Zhong, Qihuang Ding, Liang Liu, Juhua Du, Bo Tao, Dacheng The School of Computer Science National Engineering Research Center for Multimedia Software Institute of Artificial Intelligence Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University Wuhan China The School of Computer Science Faculty of Engineering The University of Sydney Australia

Sequence-to-sequence (seq2seq) learning is a popular fashion for large-scale pretraining language models. However, the previous seq2seq pretraining models generally focus on reconstructive objectives on the decoder side and neglect the effect of encoder-side supervision, which we argue may lead to sub-optimal performance. To verify our hypothesis, we first empirically study the functionalities of the encoder and decoder in seq2seq pretrained language models, and find that the encoder takes an important but under-exploitation role than the decoder regarding the downstream performance and neuron activation. Therefore, we propose an encoding-enhanced seq2seq pretraining strategy, namely E2S2, which improves the seq2seq models via integrating more efficient self-supervised information into the encoders. Specifically, E2S2 adopts two self-supervised objectives on the encoder side from two aspects: 1) locally denoising the corrupted sentence (denoising objective);and 2) globally learning better sentence representations (contrastive objective). With the help of both objectives, the encoder can effectively distinguish the noise tokens and capture high-level (i.e., syntactic and semantic) knowledge, thus strengthening the ability of seq2seq model to accurately achieve the conditional generation. On a large diversity of downstream natural language understanding and generation tasks, E2S2 dominantly improves the performance of its powerful backbone models, e.g., BART and T5. For example, upon BART backbone, we achieve +1.1% averaged gain on the general language understanding evaluation (GLUE) benchmark and +1.75% F0.5 score improvement on CoNLL2014 dataset. We also provide in-depth analyses to show the improvement stems from better linguistic representation. We hope that our work will foster future self-supervision research on seq2seq language model pretraining. Copyright © 2022, The Authors. All rights reserved.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

A Lightweight and Enhanced Semantic Segmentation Network for Mapping of Retrogressive Thaw Slumps from Sentinel-2 Images

A Lightweight and Enhanced Semantic Segmentation Network for...

引用

IEEE International Symposium on Geoscience and Remote Sensing (IGARSS)

作者： Yi Yuan Guiyun Zhou Zhonghua Su Ziyin Liu Weiwei Sun Xiangchao Meng Jie Wen Zhiqiang Wu School of Resources and Environment University of Electronic Science and Technology of China Chengdu China School of Computer and Software Engineering Xihua University Chengdu China Department of Geography and Spatial Information Techniques Ningbo University Ningbo China Faculty of Electrical Engineering and Computer Science Ningbo University Ningbo China Sichuan Provincial Public 7 Security Department Communications and Information Technology Headquarters Chengdu China

ISBN: (数字)9798350360325

ISBN: (纸本)9798350360332

Fine mapping of retrogressive thaw slumps (RTSs) holds paramount significance in the study of permafrost degradation and carbon exchange. We propose a lightweight and enhanced semantic segmentation network (LessNet) for automatically mapping the RTSs from Sentinel-2 images. LessNet is constructed on the encoder-decoder framework with innovative incorporation of attention mechanism and dual-level semantic features fusion. The lightweight architecture of LessNet eliminates the need for pre-training, and the network hyperparameters are automatically updated based on the training dataset, which allows for fast convergence of supervised learning. Experiments conducted in the Beiluhe region of the Tibetan Plateau highlight the robustness and competitive performance of the model.

关键词： Training Semantic segmentation Supervised learning Transfer learning Semantics Robustness Permafrost

来源：评论

学校读者我要写书评

暂无评论

A Novel Deep Learning-Based Method for Real-Time Face Spoof Detection

SSRN

引用

SSRN 2023年

作者： Malik, Muhammad Amir Mazhar, Tehseen Haq, Inayatul Shahzad, Tariq Ghadi, Yazeed Yaseen Mallek, Fatma Hamam, Habib Department of Computer Science and Software Engineering International Islamic Unversity Islamabad44000 Pakistan Department of Computer Science Virtual University of Pakistan Lahore55150 Pakistan School of Information Engineering Zhengzhou University Zhengzhou450001 China Department of Computer Sciences COMSATS University Islamabad Sahiwal Campus Sahiwal57000 Pakistan Department of Computer Science and Software Engineering Al Ain University Abu Dhabi15322 United Arab Emirates Faculty of Engineering University of de Moncton MonctonNBE1A3E9 Canada National College Of Business Administration & Economics Canada

This research aims to improve the detection accuracy of face spoof recognition systems by employing a hybrid model of machine learning and computer vision-based approaches. Differences, including Decision Tree, Nave Bayes, K-nearest Neighbor, Support Vector Machine, Convolutional Neural Network (CCN), and Recurrent Neural Networks are used for face spoof detection. For face spoof detection, the proposed model is a hybrid variant of the CNN-based classifier used in the proposed face spoof detection model. This study improves real-time face fake detection using machine learning and computer vision. The proposed system is based on a CNN-based classification approach with optimized hyper parameters that detect real-time face spoofing attacks using print, video, and repeat attacks, improving detection accuracy. IDIAP, USSA & and MSFD datasets are used in the simulation;the proposed model has achieved a maximum accuracy of 87.5%. a high sensitivity score of 92.45%,. © 2023, The Authors. All rights reserved.

关键词： Decision trees

来源：评论

学校读者我要写书评

暂无评论

Electrocoagulation-based AZO DYE (P4R) Removal Rate Prediction Model using Deep Learning

引用

Procedia computer science 2024年 236卷 51-58页

作者： Meryem Akoulih Smail Tigani Fouzia Byoud Meryem El Rharib Rachid Saadane Samuel Pierre Abdellah Chehri Sanae El Ghachtouli IME Lab Faculty of Science Hassan 2 University 20250 Casablanca Morocco Research and Development Unit Accsellium LLC Fez Morocco Electrical Engineering Department SIRC-LaGeS Hassania School of Public Labors 20250 Casablanca Morocco Department of Computer and Software Engineering Polytechnique Montréal Montréal Québec Canada Department of Mathematics and Computer Science Royal Military College of Canada Kingston ON 11 K7K 7B4 Canada

This article presents a study on the effectiveness of electrocoagulation (EC) for the removal of azo dyes from wastewater. The analysis was performed using a combination of statistical methods, including density estimation, correlation analysis, and deep learning for electrocoagulation performance prediction. The results showed that electrocoagulation was able to effectively remove azo dyes from the wastewater, considering the energy consumption and the mass of flocs being important factors in the process. Deep Learning (DL) is used to build our predictive model using the datasets collected during the experimentation stage. Overall, the findings suggest that electrocoagulation is a promising technique for the treatment of wastewater containing azo dyes, and that the use of statistical and machine learning methods can aid in the optimization of the process.

关键词： Electrocoagulation Wastewater Azo dye Deep learning

来源：评论

学校读者我要写书评

暂无评论

Exploiting Partial Observability and Optimized Simple State Representations in Deep Q-Learning

Exploiting Partial Observability and Optimized Simple State ...

引用

Frontiers of Information Technology (FIT)

作者： Danyaal Mahmood Usama Arshad Raja Hashim Ali Zain Ul Abideen Muhammad Huzaifa Shah Talha Ali Khan Ali Zeeshan Ijaz Nisar Ali Abu Bakar Siddique Faculty of Computer Science & Engineering GIK Institute of Engg. Sciences & Technology Topi Pakistan Dept. of Technology & Software Engineering University of Europe for Applied Sciences Berlin Germany BCMaterials Basque Center for Materials Applications & Nanostructure Spain Faculty of Electronic Systems Engineering University of Regina Regina Canada

The advent of deep Q-learning has opened up new possibilities in training autonomous agents to perform intelligently in intricate settings. This research work examines the potential of deep Q-learning in the paradigmatic Snake game, which requires an agent to navigate a grid, ingest food items, and avoid collisions. We incorporated a partially observable game state, introducing a novel level of difficulty to the task. Our proposed approach utilizes a straightforward feed-forward neural network to estimate Q-values of potential actions, given the current state. The agent progressively learns to balance between exploration and exploitation, allowing it to navigate optimally in the game scenario. Most existing studies in deep Q-learning for the Snake game overlook the complexities posed by partial observability, focusing mainly on fully observable state representations. Furthermore, the majority of the investigations delve into the efficacy of complex neural network architectures, leaving a gap in the understanding of the performance of more simplistic feed-forward networks. Our study fills this gap by offering an in-depth exploration and analysis of deep Q-learning strategies, especially when applied to partially observable game states and simple neural networks. Simulations and results reveal that adopting simpler state representations can significantly boost the agent’s performance in the Snake game. A more detailed and intricate reward function is also observed to enhance the agent’s decision-making skills and overall proficiency in the game. These observations underscore the significance of optimizing state representations and reward functions for maximizing the performance of deep Q-learning agents in the Snake game.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Prediction of Polycystic Ovary Syndrome Using Genetic Algorithm-driven Feature Selection

Prediction of Polycystic Ovary Syndrome Using Genetic Algori...

引用

IT and Industrial Technologies (ICIT), International Conference on

作者： Fatima Faridoon Raja Hashim Ali Zain Ul Abideen Nazia Shahzadi Ali Zeeshan Ijaz Usama Arshad Nisar Ali Muhammad Imad Said Nabi Faculty of Computer Science & Engineering GIK Institute of Engg. Sciences & Technology Topi Pakistan Dept. of Technology & Software Engineering University of Europe for Applied Sciences Berlin Germany BCMaterials Basque Center for Materials Applications & Nanostructure Spain Faculty of Electronic Systems Engineering University of Regina Regina Canada

PolyCystic Ovary Syndrome (PCOS) is a hormonal disorder frequently found in women of reproductive age having a significant impact on the cause of infertility. It is an endocrine condition characterized by abnormalities in female hormone levels and aberrant synthesis of male hormones. This syndrome causes ovarian malfunction, increasing the risk of miscarriage and infertility. PCOS has a wide range of symptoms, making a diagnosis difficult. In this study, we proposed a feature selection method based on genetic algorithm with logistic regression to increase the accuracy of early diagnosis. A dataset containing the clinical and biochemical characteristics of 109 patients, including 36 with PCOS, is used. The genetic algorithm is used to extract the most important features from the dataset. Then, a logistic regression model is used for classification. The proposed model outperformed the baseline model that used all features, which had an accuracy of 86.2%, and produced a markedly improved accuracy of 95.4%. These results show that the proposed model effectively locates key characteristics for PCOS diagnosis.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Optimized Flame Detection Using a YOLOv5 Neural Network

Optimized Flame Detection Using a YOLOv5 Neural Network

引用

Computing, Robotics and System sciences (ICRSS), International Conference on

作者： Shahid Karim Muhammad Ibrar Anas Bilal Akeel Qadir Irfana Bibi School of Information Engineering Xi'an Eurasia University Xi'an China Software College Shenyang Normal University Shenyang China College of Information Science and Technology Hainan Normal University Haikou China Department of Computer Science Faculty of Computing and Information Technology University of Punjab Lahore Punjab Pakistan

ISBN: (数字)9798331511296

ISBN: (纸本)9798331511302

With the rising prevalence of smart homes, there's an increasing demand for comprehensive automation solutions to mitigate fire risks, especially when homeowners are absent or in homes with elderly residents. This study focuses on enhancing flame detection techniques, particularly for smaller, hard-to-detect fire sources. A customized Yolov5 neural network was developed to improve detection precision. This enhancement was facilitated using a dataset of over 2,000 images from lab camera observations and online platforms. These images were then manually labeled and clustered for anchor dimensions using the K-means++ algorithm. The CIOU and diou _ nms techniques were integrated to optimize the network's performance. Upon training, the refined model was tested on the Jet Nano, revealing its high efficacy in flame detection.

关键词： YOLO Training Accuracy Sensitivity Service robots Neural networks Fires Clustering algorithms Smart homes Safety

来源：评论

学校读者我要写书评

暂无评论

Investigating novel machine learning based intrusion detection models for NSL-KDD data sets

Investigating novel machine learning based intrusion detecti...

引用

IT and Industrial Technologies (ICIT), International Conference on

作者： Muhammad Huzaifa Shah Muhammad Abu Bakar Raja Hashim Ali Zain Ul Abideen Usama Arshad Ali Zeeshan Ijaz Nisar Ali Muhammad Imad Said Nabi Faculty of Computer Science & Engineering GIK Institute of Engg. Sciences & Technology Topi Pakistan Dept. of Technology & Software Engineering University of Europe for Applied Sciences Berlin Germany BCMaterials Basque Center for Materials Applications & Nanostructure Spain Faculty of Electronic Systems Engineering University of Regina Regina Canada

This study investigates the application of the Mutual Information (MI) feature selection technique to improve the accuracy of Machine Learning (ML) models on NSL-KDD datasets, building upon prior research. Six ML models, namely Decision Tree (DT), Logistic Regression (LR), K-Nearest Neighbor (KNN), Random Forest (RF), Naive Bayes (NB), and Support Vector Machine (SVM) with different kernels (1st, 2nd, and 3rd), are implemented for classification purposes. The proposed DT model in this study shows higher accuracy than the DT model proposed in the original paper by Ingre et al. for Intrusion Detection System (IDS). Additionally, a multi-class classification model for NSL-KDD datasets is developed, considering both normalized and non-normalized features. Interestingly, it is observed that the models trained without normalized features achieve higher accuracies compared to those trained with normalized features. Moreover, the study enhances the classification performance of the DT-based IDS using the Correlation based Feature Selection (CFS) technique for feature selection. The proposed IDS is evaluated both before and after feature selection for multi-class classification (normal and various attack types) and binary classification (normal and abnormal data).

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Robust Methodology for Fruit Quality Prediction and Estimation Using Intelligent Learning Based Image Processing Logic

A Robust Methodology for Fruit Quality Prediction and Estima...

引用

Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES), International Conference on

作者： Malathi K Vunnava Dinesh Babu M. Ramkumar M. Ayyadurai Vivek K Hamzeh J Aljawawdeh Department of CSE Saveetha School of Engineering Saveetha Institute of Science and Technology (SIMATS) Saveetha University Chennai Department of CSE RV Institute of Technology Guntur Department of Computer Science and Engineering Rajalakshmi Engineering College Chennai Chennai Department of Mechanical Engineering Yeshwantrao Chavan College of Engineering Nagpur Department of software engineering Faculty of information technology Zarqa University Zarqa Jordan University of Business and Technology Jeddah Saudi Arabia

ISBN: (数字)9798331543624

ISBN: (纸本)9798331543631

Fruit quality prediction plays a vital role in agricultural and food industries, ensuring that only high-quality produce reaches the consumer. Traditional methods of assessing fruit quality rely on manual inspection, which is both time-consuming and prone to human error. This paper presents a robust methodology for fruit quality prediction and estimation, utilizing a hybrid deep learning model that integrates DenseN et for feature extraction with LightGBM for efficient classification. The proposed model is designed to automate the process of fruit quality assessment by analyzing key attributes such as color, texture, and shape from high-resolution fruit images. The Fruit360 dataset was used for training and testing, with images classified into categories such as fresh, ripe, overripe, and damaged. A comprehensive comparison with nine existing models, including SVM, Random Forest, and XGBoost, was conducted using performance metrics such as accuracy, precision, recall, and Fl-score. The proposed model achieved an accuracy of 96.78%, significantly outperforming all other models. Furthermore, the model's precision and recall reached 95.33% and 95.67%, respectively, demonstrating its reliability in distinguishing high-quality fruits from low-quality ones.

关键词： Training Support vector machines Accuracy Shape Computational modeling Estimation Predictive models Feature extraction Reliability Testing

来源：评论

学校读者我要写书评

暂无评论

Validating IP Reputation in Cloud Firewall Systems Using Machine Learning Driven Signature Generation and Detection Techniques

Validating IP Reputation in Cloud Firewall Systems Using Mac...

引用

IEEE Industrial Electronics and Applications Conference (IEACon)

作者： NW Chanka Lasantha Valmik Tilwari MWP Maduranga Ruvan Abeysekara Nilotpal Chakraborty Debashree Sharma Faculty of Graduate Studies IIC University of Technology Phnom Penh Cambodia Department of Electronics & Communication Engineering Indian Institute of Information Technology Guwahati India Department of Software Engineering Sri Lanka Technology Campus Padukka Sri Lanka Department of Computer Science Engineering Indian Institute of Information Technology Guwahati India

ISBN: (数字)9798350382976

ISBN: (纸本)9798350382983

This paper offers a new approach to validation for IP address reputation (IPR) using dynamic signatures achieved through the application of machine learning (ML) to conduct an accurate detection of malicious IPs. The proposed solution connects to the Amazon Web Services (AWS) Web Application Firewall (WAF) to improve security by deploying artificial intelligence (AI) models and natural language (NL) processing for automation of information extraction from IP-related information saved in the AWS object storages. All these lead to minimization of the number of false positives, real-time feedback, and optimization of IP reputation checking. The experimental outcome proves the capability of the proposed system to mitigate the IP address of the attacker and improve the protection of web applications against cybercrimes. This new approach is much advanced and a major leap ahead of other conventional IP reputation techniques, which offer a good protection shield against spam, DDoS, and other such evils.

关键词： Radio frequency Accuracy Firewalls (computing) Web services Natural languages Real-time systems IP networks Protection computer crime Optimization

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：