检索结果-内蒙古大学图书馆

9th International Conference on Computer science and Computational Intelligence, ICCSCI 2024

作者： Henry, Matthew Martianus Heryanto, Nur Adhianti Pardamean, Bens Bioinformatics and Data Science Research Center Bina Nusantara University Jakarta11480 Indonesia Computer Science Department BINUS Graduate Program Bina Nusantara University Jakarta11480 Indonesia

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. © 2024 The Authors.

关键词： Medical imaging

来源：评论

学校读者我要写书评

暂无评论

Evaluating the Effectiveness of Transformer Models for Molecular Property Prediction: A Review 9

Evaluating the Effectiveness of Transformer Models for Molec...

引用

9th International Conference on Information Technology and Digital Applications, ICITDA 2024

作者： Imani, Alyssa Pardamean, Bens Bioinformatics and Data Science Research Center Bina Nusantara University Jakarta11480 Indonesia Binus Graduate Program Bina Nusantara University Computer Science Departement Jakarta11480 Indonesia

ISBN: (纸本)9798331510732

Transformer models, originally successful in natural language processing, are now being applied to chemical and biological studies, excelling in areas such as molecular property prediction, material science, and drug discovery. BERT, a Transformer-based model, has become foundational in cheminformatics, particularly for QSAR (Quantitative Structure-Activity Relationship) modeling and ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) evaluations in drug discovery. However, achieving higher accuracy often requires designing more complex models, which can compromise their interpretability. This posing a challenge for researchers who need to understand the reasoning behind the predictions. The trade-off between accuracy and interpretability presents a critical challenge in applying black box models to real-world problems in cheminformatics. This work compares Transformer-based models with traditional machine learning and deep learning approaches, focusing on both interpretability and performance. The goal is to highlight the strengths and limitations of each method, offering insights into their optimal use in drug discovery and material science. © 2024 IEEE.

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

Improved random features for dot product kernels

The Journal of Machine Learning Research

引用

The Journal of Machine Learning Research 2024年第1期25卷 11403-11477页

作者： Jonas Wacker Motonobu Kanagawa Maurizio Filippone Data Science Department EURECOM France Statistics Program KAUST Saudi Arabia

Dot product kernels, such as polynomial and exponential (softmax) kernels, are among the most widely used kernels in machine learning, as they enable modeling the interactions between input features, which is crucial in applications like computer vision, natural language processing, and recommender systems. We make several novel contributions for improving the efficiency of random feature approximations for dot product kernels, to make these kernels more useful in large scale learning. First, we present a generalization of existing random feature approximations for polynomial kernels, such as Rademacher and Gaussian sketches and TensorSRHT, using complex-valued random features. We show empirically that the use of complex features can significantly reduce the variances of these approximations. Second, we provide a theoretical analysis for understanding the factors affecting the efficiency of various random feature approximations, by deriving closed-form expressions for their variances. These variance formulas elucidate conditions under which certain approximations (e.g., TensorSRHT) achieve lower variances than others (e.g., Rademacher sketches), and conditions under which the use of complex features leads to lower variances than real features. Third, by using these variance formulas, which can be evaluated in practice, we develop a data-driven optimization approach to improve random feature approximations for general dot product kernels, which is also applicable to the Gaussian kernel. We describe the improvements brought by these contributions with extensive experiments on a variety of tasks and datasets.

关键词： random features randomized sketches dot product kernels polynomial kernels large scale learning

来源：评论

学校读者我要写书评

暂无评论

LDA Topic Modeling for Bioinformatics Terms in arXiv Documents 9th

LDA Topic Modeling for Bioinformatics Terms in arXiv Documen...

引用

9th International Conference on Computer science and Computational Intelligence, ICCSCI 2024

作者： Karnyoto, Andrea Stevens Henry, Matthew Martianus Pardamean, Bens Bioinformatics and Data Science Research Center Bina Nusantara University Jakarta11480 Indonesia Computer Science Department BINUS Graduate Program Bina Nusantara University Jakarta11480 Indonesia

A wide variety of disciplines contribute to bioinformatics research, including computer science, biology, chemistry, mathematics, and physics. This study determines the number of research articles published on arXiv classified as bioinformatics topics and the most frequently used bioinformatics terms using topic modeling, Latent Dirichlet Allocation (LDA). An algorithm based on LDA is used to discover topics hidden within large collections of documents through the use of statistical analysis. Our research examined 226453 articles on arXiv between January 2023 and January 2024. As a result, there are more than 10521 articles categorized into bioinformatics topics. Most commonly, 6352 documents are in the "Mathematical Physics" category. The second most popular category is "Computer science," with 2950 documents. Accordingly, the terms 'RNA,' 'sequence,' 'tree,' and 'homology' are the three most commonly used terms in bioinformatics. The study of RNA plays a vital role in molecular biology;thus, the study of RNA is prevalent in bioinformatics. Sequential data refer to the order in which nucleotides or amino acids can be found in a DNA molecule or a protein. © 2024 The Authors.

关键词： arXiv bioinformatics Latent Dirichlet Allocation

来源：评论

学校读者我要写书评

暂无评论

Analysis of the 5Rs in Thailand Medication Error Classification through Natural Language Processing 20

Analysis of the 5Rs in Thailand Medication Error Classificat...

引用

20th International Joint Conference on Computer science and Software Engineering, JCSSE 2023

作者： Udomnuchaisup, Peachyasitt Imsombut, Aurawan Suwannahitatorn, Picha Saethang, Thammakorn Kasetsart University Biomedical Data Science Master's Program Faculty of Science Bangkok Thailand Kasetsart University Faculty of Science Department of Computer Science Bangkok Thailand Bangkok Thailand

ISBN: (纸本)9798350300505

Medication errors threaten patient safety considerably, underscoring the necessity for enhanced detection and prevention techniques. A prevalent classification system in hospitals relies on the standard practice of medication administration known as the Five Rights (5R). This study seeks to develop an NLP-based tool designed to expand 5R error categorization coverage and alleviate the workload of medical professionals. The proposed method focuses on Thai medical text, incorporating Thai and English vocabulary. In this investigation, we developed a supervised learning classification framework using the Universal Sentence Encoder (USE) for sentence embedding, followed by an Artificial Neural Network (ANN) for model training. Additionally, we explored a zero-shot classification model employing pre-trained Large Language Models (PLMs). Our findings reveal that the supervised learning classification model provides the most favorable performance, albeit with the limitation of reliance on labeled datasets, which can be resource intensive. Conversely, the zero-shot classification framework's performance is less optimal. However, future advancements in Thai medical PLMs may improve efficacy and present a viable alternative for medical data analysis without dependence on labeled datasets. This initiative lays the groundwork for potential future applications and advantages within Thailand's medical domain. © 2023 IEEE.

关键词： Neural networks

来源：评论

学校读者我要写书评

暂无评论

FSL-LFMG: Few-Shot Learning with Augmented Latent Features and Multitasking Generation for Enhancing Multiclass Classification on Tabular data 16th

FSL-LFMG: Few-Shot Learning with Augmented Latent Features a...

引用

16th International Joint Conference on Computational Intelligence, IJCCI 2024

作者： Nur, Aviv A. Ngan, Chun-Kit Bardeli, Rolf Data Science Program Worcester Polytechnic Institute 100 Institute Rd WorcesterMA United States thyssenkrupp Materials Services GmbH Essen Germany

ISBN: (纸本)9789897587214

In this work, we propose advancing ProtoNet that employs augmented latent features (LF) by an autoencoder and multitasking generation (MG) by STUNT in the few-shot learning (FSL) mechanism. Specifically, the achieved contributions to this work are threefold. First, we propose an FSL-LFMG framework to develop an end-to-end few-shot multiclass classification workflow on tabular data. This framework is composed of three main stages that include (i) data augmentation at the sample level utilizing autoencoders to generate augmented LF, (ii) data augmentation at the task level involving self-generating multitasks using the STUNT approach, and (iii) the learning process taking place on ProtoNet, followed by various model evaluations in our FSL mechanism. Second, due to the outlier and noise sensitivity of K-means clustering and the curse of dimensionality of Euclidean distance, we enhance and customize the STUNT approach by using K-medoids clustering that is less sensitive to noisy outliers and Manhattan distance that is the most preferable for highdimensional data. Finally, we conduct an extensive experimental study on four diverse domain datasets—Net Promoter Score segmentation, Dry Bean type, Wine type, and Forest Cover type—to prove that our FSLLFMG approach on the multiclass classification outperforms the Tree Ensemble models and the One-vs-therest classifiers by 7.8% in 1-shot and 2.5% in 5-shot learning. © 2024 by Paper published under CC license (CC BY-NC-ND 4.0)

关键词： Decision trees

来源：评论

学校读者我要写书评

暂无评论

An Evaluation Of Hospital Accreditation From The Survey With Text Vector Analysis Techniques 20

An Evaluation Of Hospital Accreditation From The Survey With...

引用

20th International Joint Conference on Computer science and Software Engineering, JCSSE 2023

作者： Prapunwattana, Dittaphong Imsombut, Aurawan Suwannahitatorn, Picha Saethang, Thammakorn Kasetsart University Biomedical Data Science Master's Program Faculty of Science Bangkok Thailand Kasetsart University Faculty of Science Department of Computer Science Bangkok Thailand Bangkok Thailand

ISBN: (纸本)9798350300505

The Healthcare Accreditation Institute has an assessment and certification process for hospitals applying for Healthcare Accreditation. The assessment process requires a large number of text-based reports. The purpose of this research was to study the text analysis of the self-assessment reports of healthcare facilities and surveyor reports on issues related to the pharmaceutical system to evaluate and rate the accreditation of medical facilities. The natural language text vector analysis technique, together with the Universal Sentence Encoder (USE) was compared to Learning Lightweight Language-agnostic Sentence Embeddings (LEALLA) for encoding data into a high-dimensional format. Next the sentence encoding feature was fed through a machine learning procedure, including artificial neural networks, logistic regression, and support vector machines to classify nursing facility accreditation ratings. The experimental results showed that the USE embedding yielded better performance than the LEALLA embedding across all models with a precision of 0.70 but took slightly longer to encode feature sentences. This research could improve the performance of the analysis and scoring. © 2023 IEEE.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

News-Driven Price Movement Forecasting with Label-Prior Graph Attention 24

News-Driven Price Movement Forecasting with Label-Prior Grap...

引用

33rd Companion of the ACM World Wide Web Conference, WWW 2023

作者： Liou, Yi-Ting Chen, Chung-Chi Huang, Hen-Hsen Chen, Hsin-Hsi Graduate Program of Data Science National Taiwan University Academia Sinica Taiwan AIST Japan Institute of Information Science Academia Sinica Taiwan Department of Computer Science and Information Engineering National Taiwan University Taiwan

ISBN: (纸本)9798400701726

This paper introduces a novel approach to stock movement prediction using multi-label classification, leveraging the interconnections between news articles and related company stocks. We present the Label-Prior Graph Attention (LPGA) model, which significantly enhances the performance of news-driven stock price movement forecasting. This model is comprised of a unique graph attention architecture, incorporating a label encoder and a text encoder, designed to effectively capture and utilize the relationships between labels in a graph-based context. Our model demonstrates superior performance over several benchmark models. The LPGA model’s efficacy is further validated through experiments on two multilabel datasets. The model outperforms established baseline models across various evaluation metrics. The success of the LPGA model in both stock movement prediction and general multi-label classification tasks indicates its potential as a versatile tool in the realm of machine learning and financial analysis. © 2024 Copyright held by the owner/author(s). Publication rights licensed to ACM.

关键词： Graphic methods

来源：评论

学校读者我要写书评

暂无评论

Prior-Guided Parallel Residual Bi-Fusion Network in USV Obstacle Detection

Prior-Guided Parallel Residual Bi-Fusion Network in USV Obst...

引用

2023 International Conference on Consumer Electronics - Taiwan, ICCE-Taiwan 2023

作者： Hsu, Chih-Chung Yang, Sophia Hou, Xiu-Yu Jhang, Yu-An Institute of Data Science National Cheng Kung University Taiwan National Tsing Hua University Interdisciplinary Program of Electrical Engineering and Computer Science Taiwan

ISBN: (纸本)9798350324174

In this paper, we propose a novel Prior-Guided Parallel Residual Bi-Fusion Feature Pyramid Network (PPRB-FPN) for accurate obstacle detection in unmanned surface vehicle (USV) sailing. Our method tackles the challenge of detecting small objects, which are prone to information vanishing. To the end, we leverage the PRB-FPN for small object detection and YOLOv7 as a single-stage object detector to effectively identify obstacles. Our experimental results on the Obstacle Detection Challenge dataset at the 1st Workshop on Maritime Computer Vision (MaCVi) demonstrate that our method outperforms both Mask R-CNN (mrcnn) and YOLOv7, achieving an F-avg score of 0.514. © 2023 IEEE.

关键词： Unmanned surface vehicles

来源：评论

学校读者我要写书评

暂无评论

Bioinformatics exploration into potential mechanisms of Calamus draco for colorectal cancer treatment 9th

Bioinformatics exploration into potential mechanisms of Cala...

引用

9th International Conference on Computer science and Computational Intelligence, ICCSCI 2024

作者： Kenneth, Christian Nirwantono, Rudi Pardamean, Bens Bioinformatics and Data Science Research Center Bina Nusantara University Jakarta11480 Indonesia Biotechnology Department Faculty of Engineering Bina Nusantara University Jakarta11480 Indonesia Computer Science Department BINUS Graduate Program Computer Science Program Bina Nusantara University Jakarta11480 Indonesia

As one of the cancer types with the highest incidence rates, colorectal cancer (CRC) would benefit from treatments with fewer side effects and reduced treatment-resistant potential. One of the options is to harness the anti-CRC potential of natural products. Previous studies have shown that Calamus draco exudate, dragon's blood, has anticancer activity in liver cancer and acute myeloid leukemia, but its bioactivity has not been studied in CRC. Here we conduct a bioinformatics study based on network pharmacology to explore the anti-CRC potential and mechanism of C. draco-derived compounds. The bioinformatics pipeline is composed of compound and target collection, biological network evaluation, and enrichment analysis. We found that there are 43 bioactive compounds from C. draco targeting 91 CRC-related targets, of which most compounds target MEN1, PTGS2, and IDH1. Further analyses show that the targets of C. draco are involved in the cellular response to hypoxia. By inhibiting those targets, C. draco bioactive compounds can potentially hinder angiogenesis and increase treatment response efficacy. © 2024 The Authors.

关键词： Lung cancer

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：