检索结果-内蒙古大学图书馆

Surface phase diagram of CsSnI3 from first-principles calculations

Physical Review Materials 2024年第9期8卷 093401页

作者： Kejia Li Chadawan Khamdang Mengen Wang Department of Electrical and Computer Engineering Materials Science and Engineering Program

CsSnI3 is widely studied as an environmentally friendly Pb-free perovskite material for optoelectronic device applications. To further improve material and device performance, it is important to understand the surface structures of CsSnI3. We generate surface structures with various stoichiometries, perform density functional theory calculations to create phase diagrams of the CsSnI3 (001), (110), and (100) surfaces, and determine the most stable surfaces under a wide range of Cs, Sn, and I chemical potentials. Under I-rich conditions, surfaces with Cs vacancies are stable, which lead to partially occupied surface states above the valence band maximum. Under I-poor conditions, we find the stoichiometric (100) surface to be stable under a wide region of the phase diagram, which does not have any surface states and can contribute to long charge-carrier lifetimes. Consequently, the I-poor (Sn-rich) conditions will be more beneficial to improve the device performance.

关键词： Electronic structure Surface reconstruction Surface states Perovskites Semiconductors Density functional theory

来源：评论

学校读者我要写书评

暂无评论

Exploring EfficientNet Variants for Image Encoding in Auxiliary Signal Guided Knowledge Encoder-Decoder Framework

引用

Procedia computer science 2024年 245卷 391-398页

作者： Matthew Martianus Henry Nur Adhianti Heryanto Bens Pardamean Bioinformatics and Data Science Research Center Bina Nusantara University Jakarta Indonesia 11480 Computer Science Department BINUS Graduate Program - Master of Computer Science Bina Nusantara University Jakarta Indonesia 11480

The growing number of medical images has led to radiologist burnout, which seriously impacts the radiologist's performance. To address the previously mentioned issue, an Auxiliary Signal Guided Knowledge (ASGK) multimodal encoder-decoder framework was designed to automatically generate the medical report based on the proposed medical graph and natural language decoder. It utilizes DenseNet-121 as the image encoder. With DenseNet-121 lack of computational and memory efficiency, this study aims to explore the potential of EfficientNetB0 to EfficientNetB4 as an ASGK image encoder substitute. The framework is trained with IU X-Ray dataset for 30 epochs, with Adam optimizer, a learning rate of 0.01 with 0.8 decay rate, binary cross entropy loss for the medical tags, and cross-entropy loss for the generated medical captions. During the framework training process with each image encoder, the parameter that achieves the highest CIDEr score on the validation set is considered the best image encoder parameter and will be used on the test set. On the test set, EfficientNetB3 as an ASGK image encoder has been shown to increase the CIDEr score to 0.35, a significant increase from the 0.28 CIDEr score obtained by the ASGK using DenseNet-121. This score is only a 1% decrease from the best validation score. It suggests that not only EfficientNetB3 increases the framework's performance, it is also less prone to overfitting. This study has demonstrated that EfficientNetB3 is a potential image encoder substitute for DenseNet-121 in the ASGK framework.

关键词： Automated medical report generation DenseNet-121 EfficientNet Encoder-decoder framework

来源：评论

学校读者我要写书评

暂无评论

Imitation Learning from Purified Demonstrations 41

Imitation Learning from Purified Demonstrations

引用

41st International Conference on Machine Learning, ICML 2024

作者： Wang, Yunke Dong, Minjing Zhao, Yukun Du, Bo Xu, Chang School of Computer Science National Engineering Research Center for Multimedia Software Institute of Artificial Intelligence Wuhan Institute of Data Intelligence Wuhan University China Department of Computer Science City University of Hong Kong Hong Kong School of Computer Science Faculty of Engineering The University of Sydney Australia

Imitation learning has emerged as a promising approach for addressing sequential decision-making problems, with the assumption that expert demonstrations are optimal. However, in real-world scenarios, most demonstrations are often imperfect, leading to challenges in the effectiveness of imitation learning. While existing research has focused on optimizing with imperfect demonstrations, the training typically requires a certain proportion of optimal demonstrations to guarantee performance. To tackle these problems, we propose to purify the potential noises in imperfect demonstrations first, and subsequently conduct imitation learning from these purified demonstrations. Motivated by the success of diffusion model, we introduce a two-step purification via diffusion process. In the first step, we apply a forward diffusion process to smooth potential noises in imperfect demonstrations by introducing additional noise. Subsequently, a reverse generative process is utilized to recover the optimal demonstration from the diffused ones. We provide theoretical evidence supporting our approach, demonstrating that the distance between the purified and optimal demonstration can be bounded. Empirical results on MuJoCo and RoboSuite demonstrate the effectiveness of our method from different aspects. Copyright 2024 by the author(s)

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Security Challenges and Issues in Cloud Computing

Security Challenges and Issues in Cloud Computing

引用

International Symposium on Networks, computers and Communications (ISNCC)

作者： Rika Zakiyyah Daniel Permana Devan Valenecio Andry Chowanda Yohan Muliono Zahra Nabila Izdihar Computer Science Department Cyber Security Program School of Computer Science Jakarta Indonesia Computer Science Department Computer Science Program School of Computer Science Jakarta Indonesia

ISBN: (数字)9798350364910

ISBN: (纸本)9798350364927

Managing data has changed significantly because of cloud computing, which offers scalabe, flexible and reasonably priced solutions to enterprises and to people as well such as Amazon, Google, and Microsoft expanding their infrastructure to meet the demand from the customer. With such advantages there will also be disadvantages such as security problems. This paper will discuss about security problems such as encryption, access control, multi-factor authentication, regular audits, secure configurations, incident response planning. This paper will also talk about security measures to deal with threats such as malicious insiders, data abuse, unsecured interfaces and APIs, shared technology complications, data loss or leakage, hijacking, and enigmatic risk profiles. This study examines and evaluates a range of cloud computing security concerns and issues through a systematic analysis of the literature. The goal of the research is to raise public awareness of the need for cloud computing security and to offer possible remedies. Identify and classify network security threats and vulnerabilities in cloud computing by doing an extensive examination of the body of research and empirical investigations is one of the answer to the questions from the paper. In the end, this take a look at unearths vulnerabilities and their practical implications through very well reading network safety in cloud computing.

关键词： Training Systematics Multi-factor authentication Navigation Cloud computing security Network security Loss measurement Encryption Planning Protection

来源：评论

学校读者我要写书评

暂无评论

Sign Language to Sentence Formation:A Real Time Solution for Deaf People

引用

computers, Materials & Continua 2022年第8期72卷 2501-2519页

作者： Muhammad Sanaullah Muhammad Kashif Babar Ahmad Tauqeer Safdar Mehdi Hassan Mohd Hilmi Hasan Amir Haider Bahauddin Zakariya University Department of Computer ScienceMultan60000Pakistan Air University Department of Computer ScienceMultan60000Pakistan Air University Department of Computer ScienceIslamabad44000Pakistan Centre for Research in Data Science Department of Computer and Information SciencesUniversiti Teknologi PETRONASSeri Iskandar32610PerakMalaysia Department of Intelligent Mechatronics Engineering Sejong UniversitySeoul05006Korea

Communication is a basic need of every human being to exchange thoughts and interact with the *** peoples usually confab through different spoken languages,whereas deaf people cannot do ***,the Sign Language(SL)is the communication medium of such people for their conversation and interaction with the *** SL is expressed in terms of specific gesture for every word and a gesture is consisted in a sequence of performed *** acute people normally observe these signs to understand the difference between single and multiple gestures for singular and plural words *** signs for singular words such as I,eat,drink,home are unalike the plural words as school,cars,players.A special training is required to gain the sufficient knowledge and practice so that people can differentiate and understand every gesture/sign *** researches have been performed to articulate the computer-based solution to understand the single gesture with the help of a single hand *** complete understanding of such communications are possible only with the help of this differentiation of gestures in computer-based solution of SL to cope with the real world ***,there is still a demand for specific environment to automate such a communication solution to interact with such type of special *** research focuses on facilitating the deaf community by capturing the gestures in video format and then mapping and differentiating as single or multiple gestures used in ***,these are converted into the respective words/sentences within a reasonable *** provide a real time solution for the deaf people to communicate and interact with the society.

关键词： Sign language machine learning conventional neural network image processing deaf community

来源：评论

学校读者我要写书评

暂无评论

Analysis of Bitcoin Prediction using Transparent and Opaque Machine Learning Algorithms 15

Analysis of Bitcoin Prediction using Transparent and Opaque ...

引用

15th International Conference on Advances in Computing, Control, and Telecommunication Technologies, ACT 2024

作者： Singh, Kavita R. Jamar, Anushri V. Khadgi, Pracheta P. Kapse, Smita Kumar, Ajay Department of Computer Technology Yeshwantrao Chavan College of Engineering Maharashtra Nagpur India School of Big Data and Computer Science Guizhou Province Guiyang City China

ISBN: (纸本)9798331300579

The unpredictable nature of cryptocurrency markets, particularly Bitcoin, has attracted significant attention from investors, researchers, and financial institutions seeking to understand and predict price movements. This study aims to analyse the effectiveness of both transparent and opaque machine learning algorithms in predicting Bitcoin prices. Transparent algorithms, such as linear regression and decision trees, provide clear insights into the factors influencing predictions. On the other hand, opaque algorithms, like ensemble methods, offer complex, non-linear models that may capture intricate patterns in the data. The main analysis is to lower the mean absolute error and enhance the r2 square, both key indicators of forecast accuracy. The study presents that linear regression and random forest outperformed others, achieving the smallest error and highest variance explanation, suggesting their effectiveness in Bitcoin price prediction with and without outliers respectively. In contrast, the support vector machine underperforms compared to all other machine learning algorithms under consideration. This paper sheds light on the most capable machine-learning techniques for Bitcoin forecasting, aiding trader investment choices and guiding academic study in the field of prediction. © Grenze Scientific Society, 2024.

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

Semi-automated meningioma segmentation with bounding boxes

引用

Procedia computer science 2024年 245卷 583-590页

作者： Nur Adhianti Heryanto Mahmud Isnan Matthew Martianus Henry Bens Pardamean Bioinformatics and Data Science Research Center Bina Nusantara University Jakarta Indonesia 11480 Computer Science Department BINUS Graduate Program - Master of Computer Science Bina Nusantara University Jakarta Indonesia 11480

Segmentation is manually performed by physicians, which takes considerable time and may be subject to observers. Automating this task can increase efficiency and consistency. Existing studies on meningioma segmentation used data from limited study centers, indicating the need for research on multi-center data to assess generalizability. In this work, two semi-automated methods with bounding box priors, LiteMedSAM and BBU-Net, are evaluated on the brain tumor segmentation (BraTS) 2023 meningioma dataset collected from five study-centers. Preprocessing included exclusion of small tumors, z-score normalization, and extraction of slices that contain tumors, generating 25,602 2D axial magnetic resonance imaging (MRI) scans. A fine-tuning strategy is adopted for LiteMedSAM while BBU-Net is trained from scratch. The models are evaluated using a five-fold cross-validation, with data split at the case level. Results show that while U-Net models can achieve performance close to LiteMedSAM, the foundation model has overall better performance, with more than 90% in all evaluation scores.

关键词： Bounding boxes Meningioma Semi-automated segmentation

来源：评论

学校读者我要写书评

暂无评论

Stabilizing Adversarial Training for Generative Networks

Stabilizing Adversarial Training for Generative Networks

引用

2023 IEEE International Conference on Big data, Bigdata 2023

作者： Gerych, Walter Hickey, Kevin Hartvigsen, Thomas Buquicchio, Luke Alajaji, Abdulaziz Chandrasekaran, Kavin Mansoor, Hamid Agu, Emmanuel Rundensteiner, Elke University of Virginia School of Data Science CharlottesvilleVA United States King Saud University College of Computer and Informational Sciences Riyadh Saudi Arabia University of Victoria Department of Computer Science VictoriaBC Canada Worcester Polytechnic Institute Department of Data Science WorcesterMA United States Worcester Polytechnic Institute Department of Computer Science WorcesterMA United States

ISBN: (纸本)9798350324457

Generative modeling is a powerful technique that involves creating machine learning models capable of creating new data similar to the data it was trained on. Generative Adversarial Networks (GANs) are a leading approach for generative modeling. However, GAN training is known to be a notoriously difficult task. GAN convergence issues are largely caused by the supports of the real and generated distributions being disjoint. To tackle this open problem, we propose a novel GAN pre-training process that effectively aligns the supports of the generated and real data prior to applying traditional adversarial GAN training. The key component of our method, called AlignGAN, is learning a mapping between the input data distribution and a latent representation defined over a hypersphere, regularized by a One Class Classifier. This successfully encourages the generator to produce samples throughout the support of the real data, while not generating samples outside the support. We maintain support alignment through low-bandwidth noise convolutions and additional One Class regularization, leading to continued stable GAN training. We validate our approach against leading stabilization methods on three benchmark datasets, showing AlignGAN routinely produces the best results. © 2023 IEEE.

关键词： distribution GAN support alignment

来源：评论

学校读者我要写书评

暂无评论

Enhancing face recognition performance: a comprehensive evaluation of deep learning models and a novel ensemble approach with hyperparameter tuning

引用

Soft Computing 2024年第20期28卷 12399-12424页

作者： Selvaganesan, Jana Sudharani, B. Shekhar, S. N. Chandra Vaishnavi, K. Priyadarsini, K. Raju, K. Srujan Rao, T. Srinivasa Department of Electronics and Communication Engineering Vel Tech Rangarajan Dr. Sagunthala R & ampD Institute of Science and Technology Avadi Tamil Nadu Chennai600062 India Guntur Dist. Andhra Pradesh Vadlamudi India Department of Electronics and Communication Engineering Sreenidhi Institute of Science and Technology Telangana Hyderabad501301 India Department of Computer Science and Engineering Sona College of Technology Tamil Nadu Salem636005 India Department of Data Science and Business Systems School of Computing Faculty of Engineering and Technology SRM Institute of Science and Technology Tamil Nadu Kattankulathur603203 India Department of Computer Science and Engineering CMR Technical Campus Medchal Telangana Hyderabad501401 India Department of Electronics and Communication Engineering Aditya Engineering College Surampalem533437 India

In response to growing security concerns and the increasing demand for face recognition (FR) technology in various sectors, this research explores the application of deep learning techniques, specifically pre-trained Convolutional Neural Network (CNN) models, in the field of FR. The study harnesses the power of five pre-trained CNN models—DenseNet201, ResNet152V2, MobileNetV2, SeResNeXt, and Xception—for robust feature extraction, followed by SoftMax classification. A novel weighted average ensemble model, meticulously optimized through a grid search technique, is introduced to augment feature extraction and classification efficacy. Emphasizing the significance of robust data pre-processing, encompassing resizing, data augmentation, splitting, and normalization, the research endeavors to fortify the reliability of FR systems. Methodologically, the study systematically investigates hyperparameters across deep learning models, fine-tuning network depth, learning rate, activation functions, and optimization methods. Comprehensive evaluations unfold across diverse datasets to discern the effectiveness of the proposed models. Key contributions of this work encompass the utilization of pre-trained CNN models for feature extraction, extensive evaluation across multiple datasets, the introduction of a weighted average ensemble model, emphasis on robust data pre-processing, systematic hyperparameter tuning, and the utilization of comprehensive evaluation metrics. The results, meticulously analyzed, unveil the superior performance of the proposed method, consistently outshining alternative models across pivotal metrics, including Recall, Precision, F1 Score, Matthews Correlation Coefficient (MCC), and Accuracy. Notably, the proposed method attains an exceptional accuracy of 99.48% on the labeled faces in the wild (LFW) dataset, surpassing erstwhile state-of-the-art benchmarks. This research represents a significant stride in FR technology, furnishing a dependable and accurate

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Toward Identifying The Best Base Classifier in Multi Label Classification-an Investigative Study 24

Toward Identifying The Best Base Classifier in Multi Label C...

引用

24th International Arab Conference on Information Technology, ACIT 2023

作者： Alzyoud, Mazen Alazaidah, Raed Alzoubi, Haneen Al-Shanableh, Najah Aljaidi, Mohammad Almatarneh, Sattam Al Al-Bayu University Department of Computer Science Mafraq Jordan Zarqa University Department of Data Science and Ai Zarqa Jordan Zarqa University Department of Computer Science Zarqa Jordan Zarqa University Department of Artificial Intelligence Zarqa Jordan

ISBN: (纸本)9798350384307

Classification is a significant task in data mining, machine learning and data science. It aims to predict the class label for a new case accurately. Classification is of two types: Single Label Classification (SLC) and Multi-label Classification (MLC). In SLC, cases and instances are linked to one class label only, while in MLC, cases and instances could be linked to more than once class labels. Several SLC algorithms have been adapted and upgraded to handle MLC and showed different predictive performance. Hence, this paper attempts to identify and investigate the best SLC that could handle the problem of MLC with respect to three datasets and using three evaluation metrics. Moreover, the paper also aims to identify the best Problem Transformation Method (PTM) among five well-known methods. The results revealed that RandomForest and DecisionTable showed the best performance among the fifteen classifier. Also, Most Frequent Label (MFL) is the best transformation method among the five considered PTMs in term of Accuracy, Precision, and Recall. © 2023 IEEE.

关键词： Algorithms classification data mining data science machine learning multi-label classification

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：