检索结果-内蒙古大学图书馆

Image Classification Model Using Deep Learning on the Edge Device 1

学校读者我要写书评

暂无评论

3rd International Conference on Innovative Computing, IC 2020

作者： Kristiani, Endah Yang, Chao-Tung Huang, Chin-Yin Chan, Yu-Wei Fathoni, Halim Department of Industrial Engineering and Enterprise Information Tunghai University Taichung407302 Taiwan Department of Informatics Krida Wacana Christian University Jakarta11470 Indonesia Department of Computer Science Tunghai University Taichung407302 Taiwan College of Computing and Informatics Providence University Taichung43301 Taiwan

ISBN: (数字)9789811559594

ISBN: (纸本)9789811559587

Training and inference are a crucial stage of deep learning. For better performance, it is essential to use an optimized approach during the training stage, as the selection of a system, network, model, and fine-tuning the training and hyperparameters. In this paper, an optimized model of the deep learning training and inference is proposed using insects dataset. The dataset consists of 613 images belonging to 10 classes of insects. In order to accelerate the deep learning training model, this system implemented CPU optimization and hyperparameter tuning. TensorFlow was applied as a framework for the training model with InceptionV3 as the topology. The evaluation of the model is demonstrated by the confusion matrix, precision-recall curve, and receiver operating characteristic (ROC). On the edge part, the model is optimized by the model optimizer technique for a better prediction result and speed. Finally, a Raspberry Pi 4 is used to inference the predictive model as an edge device. © 2020, Springer Nature Singapore Pte Ltd.

关键词： Image classification

Brain multigraph prediction using topology-aware adversarial graph neural network

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Bessadok, Alaa Mahjoub, Mohamed Ali Rekik, Islem BASIRA Lab Faculty of Computer and Informatics Istanbul Technical University Istanbul Turkey Higher Institute of Informatics and Communication Technologies University of Sousse Tunisia National Engineering School of Sousse University of Sousse LATIS- Laboratory of Advanced Technology and Intelligent Systems Sousse4023 Tunisia School of Science and Engineering Computing University of Dundee United Kingdom

Brain graphs (i.e, connectomes) constructed from medical scans such as magnetic resonance imaging (MRI) have become increasingly important tools to characterize the abnormal changes in the human brain. Due to the high acquisition cost and processing time of multimodal MRI, existing deep learning frameworks based on Generative Adversarial Network (GAN) focused on predicting the missing multimodal medical images from a few existing modalities. While brain graphs help better understand how a particular disorder can change the connectional facets of the brain, synthesizing a target brain multigraph (i.e, multiple brain graphs) from a single source brain graph is strikingly lacking. Additionally, existing graph generation works mainly learn one model for each target domain which limits their scalability in jointly predicting multiple target domains. Besides, while they consider the global topological scale of a graph (i.e., graph connectivity structure), they overlook the local topology at the node scale (e.g., how central a node is in the graph). To address these limitations, we introduce topology-aware graph GAN architecture (topoGAN), which jointly predicts multiple brain graphs from a single brain graph while preserving the topological structure of each target graph. Its three key innovations are: (i) designing a novel graph adversarial auto-encoder for predicting multiple brain graphs from a single one, (ii) clustering the encoded source graphs in order to handle the mode collapse issue of GAN and proposing a cluster-specific decoder, (iii) introducing a topological loss to force the prediction of topologically sound target brain graphs. The experimental results using five target domains demonstrated the outperformance of our method in brain multigraph prediction from a single graph in comparison with baseline approaches. © 2021, CC BY-NC-SA.

关键词： Graphic methods

Application of Graph-Curvature Features in Auxiliary Diagnosis for Histopathological Image Identification of Gastric Cancer

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： He, Ruilin Li, Chen Yang, Xinyi Yang, Jinzhu Jiang, Tao Grzegorzek, Marcin Sun, Hongzan Microscopic Image and Medical Image Analysis Group College of Medicine and Biological Information Engineering Northeastern University Shenyang China Key Laboratory of Intelligent Computing in Medical Image Ministry of Education Northeastern University Liaoning Shenyang China School of Intelligent Medicine Chengdu University of Traditional Chinese Medicine Chengdu China International Joint Institute of Robotics and Intelligent Systems Chengdu University of Information Technology Chengdu China Institute for Medical Informatics University of Luebeck Luebeck Germany Department of Knowledge Engineering University of Economics in Katowice Katowice Poland Shengjing Hospital of China Medical University Shenyang China School of Computer Science and Engineering

Background: Histopathology diagnosis is often regarded as the final diagnostic method for malignant tumors, but it has some drawbacks. This paper explores a computer-aided diagnostic method that can identify benign and malignant gastric cancer with histopathology images. Method: This article obtains the most suitable process through multiple experiments, compared multiple methods and features for classification. Firstly, the U-net is applied to segment the image. Next, the nucleus is extracted from the segmented image and the Minimum Spanning Tree (MST) diagram structure is drawn. The third step is to extract the graph-curvature features of histopathology image according to the MST image. Finally, by inputting graph-curvature features into the classifier, the recognition results for benign or malignant can be obtained. Result: During the experiment, we use various methods for comparison. In the image segmentation stage, U-net, watershed algorithm and Otsu threshold segmentation methods are used respectively. Combined with multiple indicators, we find that the U-net method is the most suitable for segmentation of histopathology images. In the feature extraction stage, in addition to extracting graph-edge feature and graph-curvature feature, several basic image features are also extracted, including Red, Green, Blue feature, Gray-Level Co-occurrence Matrix feature, Histogram of Oriented Gradient feature, and Local Binary Pattern feature. In the classifier design stage, we experimented with various methods, such as Support vector machine (SVM), Random forest, Artificial Neural Network, K Nearest Neighbors, VGG-16 and Inception-V3. Through the comparison and analysis, the classification results with an accuracy of 98.57% can be obtained by inputting the graph-curvature feature into SVM classifier. Conclusion: This paper has created a unique feature, graph-curvature feature based on MST to represent and analyze histopathology images. This graph-based feature can be used

关键词： Local binary pattern

Quantum architecture search with meta-learning

学校读者我要写书评

暂无评论

arXiv 2021年

作者： He, Zhimin Chen, Chuangtao Li, Lvzhou Zheng, Shenggen Situ, Haozhen School of Electronic and Information Engineering Foshan University Foshan528000 China School of Mechatronic Engineering and Automation Foshan University Foshan528000 China Institute of Quantum Computing and Computer Science Theory School of Computer Science and Engineering Sun Yat-Sen University Guangzhou510006 China Peng Cheng Laboratory Shenzhen518055 China College of Mathematics and Informatics South China Agricultural University Guangzhou510642 China

Variational quantum algorithms (VQAs) have been successfully applied to quantum approximate optimization algorithms, variational quantum compiling and quantum machine learning models. The performances of VQAs largely depend on the architecture of parameterized quantum circuits (PQCs). Quantum architecture search (QAS) aims to automate the design of PQCs in different VQAs with classical optimization algorithms. However, current QAS algorithms do not use prior experiences and search the quantum architecture from scratch for each new task, which is inefficient and time consuming. In this paper, we propose a meta quantum architecture search (MetaQAS) algorithm, which learns good initialization heuristics of the architecture (i.e., meta-architecture), along with the meta-parameters of quantum gates from a number of training tasks such that they can adapt to new tasks with a small number of gradient updates, which leads to fast learning on new tasks. The proposed MetaQAS can be used with arbitrary gradient-based QAS algorithms. Simulation results of variational quantum compiling on three- and four-qubit circuits show that the architectures optimized by MetaQAS converge much faster than a state-of-the-art gradient-based QAS algorithm, namely DQAS. MetaQAS also achieves a better solution than DQAS after fine-tuning of gate parameters. Copyright © 2021, The Authors. All rights reserved.

关键词： Learning algorithms

On the exploitation of neuroevolutionary information: Analyzing the past for a more efficient future

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Garciarena, Unai Lourenço, Nuno Machado, Penousal Santana, Roberto Mendiburu, Alexander Department of Computer Science and Artificial Intelligence University of the Basque Country Donostia-San Sebastian Spain CISUC Department of Informatics Engineering Coimbra Portugal Department of Computer Architecture and Technology University of the Basque Country Donostia-San Sebastian Spain

Neuroevolutionary algorithms, automatic searches of neural network structures by means of evolutionary techniques, are computationally costly procedures. In spite of this, due to the great performance provided by the architectures which are found, these methods are widely applied. The final outcome of neuroevolutionary processes is the best structure found during the search, and the rest of the procedure is commonly omitted in the literature. However, a good amount of residual information consisting of valuable knowledge that can be extracted is also produced during these searches. In this paper, we propose an approach that extracts this information from neuroevolutionary runs, and use it to build a metamodel that could positively impact future neural architecture searches. More specifically, by inspecting the best structures found during neuroevolutionary searches of generative adversarial networks with varying characteristics (e.g., based on dense or convolutional layers), we propose a Bayesian network-based model which can be used to either find strong neural structures right away, conveniently initialize different structural searches for different problems, or help future optimization of structures of any type to keep finding increasingly better structures where uninformed methods get stuck into local optima. © 2021, CC BY.

关键词： Generative adversarial networks

Understanding the brain with attention: A survey of transformers in brain sciences

学校读者我要写书评

暂无评论

Brain-X

Brain-X 2023年第3期1卷

作者： Chen, Cheng Wang, Huilin Chen, Yunqing Yin, Zihan Yang, Xinye Ning, Huansheng Zhang, Qian Li, Weiguang Xiao, Ruoxiu Zhao, Jizong School of Computer and Communication Engineering University of Science and Technology Beijing Beijing China Department of Neurosurgery Beijing Tiantan Hospital Capital Medical University Beijing China China National Clinical Research Center for Neurological Diseases Beijing China Department of Health Technology and Informatics Hong Kong Polytechnic University Hong Kong Shunde Innovation School University of Science and Technology Beijing Guangdong Foshan China

Owing to their superior capabilities and advanced achievements, Transformers have gradually attracted attention with regard to understanding complex brain processing mechanisms. This study aims to comprehensively review and discuss the applications of Transformers in brain sciences. First, we present a brief introduction of the critical architecture of Transformers. Then, we overview and analyze their most relevant applications in brain sciences, including brain disease diagnosis, brain age prediction, brain anomaly detection, semantic segmentation, multi-modal registration, functional Magnetic Resonance Imaging (fMRI) modeling, Electroencephalogram (EEG) processing, and multi-task collaboration. We organize the model details and open sources for reference and replication. In addition, we discuss the quantitative assessments, model complexity, and optimization of Transformers, which are topics of great concern in the field. Finally, we explore possible future challenges and opportunities, exploiting some concrete and recent cases to provoke discussion and innovation. We hope that this review will stimulate interest in further research on Transformers in the context of brain sciences. © 2023 The Authors. Brain-X published by John Wiley & Sons Australia, Ltd on behalf of Ainuohui Medical Technology.

关键词： brain science EEG fMRI MRI transformer

Supervised classification using graph-based space partitioning for multiclass problems 25

学校读者我要写书评

暂无评论

Supervised classification using graph-based space partitioni...

25th International Conference on Pattern Recognition, ICPR 2020

作者： Yanev, Nicola Valev, Ventzeslav Suliman, Karima Ben Krzyzak, Adam Institute of Mathematics and Informatics Bulgarian Academy of Sciences Sofia1113 Bulgaria Department of Computer Science and Software Engineering Concordia University MontrealQCH3G 1M8 Canada

ISBN: (纸本)9781728188089

We introduce and investigate in multiclass setting an efficient classifier which partitions the training data by means of multidimensional parallelepipeds called boxes. We show that multiclass classification problem at hand can be solved by combining the heuristic minimum clique cover approach and the k-nearest neighbor rule. Our algorithm is motivated by an algorithm for partitioning a graph into a minimal number of cliques. The main advantage of a new classifier called Box classifier is that it optimally utilizes the geometrical structure of the training set by reducing the l-class classification problem to a single nearest neighbor problem. We discuss computational complexity of the proposed Box classifier. The extensive experiments performed on the simulated and real data from UCI Machine Learning Repository for binary and multiclass problems show that in almost all cases the Box classifier performs significantly better than k-NN, SVM and decision trees. © 2020 IEEE

关键词： Decision trees

HPTMT parallel operators for high performance data science & data engineering

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Abeykoon, Vibhatha Kamburugamuve, Supun Widanage, Chathura Perera, Niranda Uyar, Ahmet Kanewala, Thejaka Amila Von Laszewski, Gregor Fox, Geoffrey Luddy School of Informatics Computing and Engineering BloomingtonIN47408 United States Biocomplexity Institute Initiative and Computer Science Dept. University of Virginia Indiana University Alumni IN47408 United States

Data-intensive applications are becoming commonplace in all science disciplines. They are comprised of a rich set of sub-domains such as data engineering, deep learning, and machine learning. These applications are built around efficient data abstractions and operators that suit the applications of different domains. Often lack of a clear definition of data structures and operators in the field has led to other implementations that do not work well together. The HPTMT architecture that we proposed recently, identifies a set of data structures, operators, and an execution model for creating rich data applications that links all aspects of data engineering and data science together efficiently. This paper elaborates and illustrates this architecture using an end-to-end application with deep learning and data engineering parts working together. © 2021, CC BY.

关键词： Data structures

Interactive Visualization of Deep Learning for 3D Brain Data Analysis

学校读者我要写书评

暂无评论

Interactive Visualization of Deep Learning for 3D Brain Data...

IEEE International Conference on Cognitive informatics

作者： Huang Li Shiaofen Fang Joaquí n Goñ i Andrew J. Saykin Li Shen Department of Computer &amp Information Science Purdue School of Science Indiana University Purdue University Indianapolis School of Industrial Engineering &amp Weldon School of Biomedical Engineering Purdue University Center for Neuroimaging School of Medicine Indiana University Department of Biostatistics Epidemiology and Informatics Perelman School of Medicine University of Pennsylvania

With multiple hidden layers and massive combinations of features and weights, deep learning models are hard to understand, and even more difficult to interact with. In this paper we describe a visual analytics platform to help with the understanding of and interaction with the deep learning process of human brain image data. A brain connectome network dataset is used to train a classifier for the diagnosis of Alzheimer's Disease (AD). 3D rendering of brain images is integrated into the interactive visualization process of a deep neural network to bring contextual information of the application to the analysis framework. A backpropagation algorithm is applied to track the image features that are captured by each node in the hidden layers. Our results demonstrate that interactive visualization can not only help the understanding of the deep learning process, but also provide a platform for domain experts to interact with and assist in the learning process, which can potentially enhance the interpretability and accuracy of the analysis.

关键词：