Multi-model ensemble with rich spatial information for object detection
اثر گروهی چند مدلی با اطلاعات مکانی غنی برای ردیابی شی-2020
Due to the development of deep learning networks and big data dimensionality, research on ensemble deep learning is receiving an increasing amount of attention. This paper takes the object detection task as the research domain and proposes an object detection framework based on ensemble deep learning. To guarantee the accuracy as well as real-time detection, the detector uses a Single Shot MultiBox Detector (SSD) as the backbone and combines ensemble learning with context modeling and multi-scale feature representation. Two modes were designed in order to achieve ensemble learning: NMS Ensembling and Feature Ensembling. In addition, to obtain contextual information, we used dilated convolution to ex- pand the receptive field of the network. Compared with state-of-the-art detectors, our detector achieves superior performance on the PASCAL VOC set and the MS COCO set.
Keywords: Ensemble learning | Object detection | Dilated convolution | Feature fusion
A non-canonical hybrid metaheuristic approach to adaptive data stream classification
یک روش متاوریستی ترکیبی غیر متعارف برای طبقه بندی جریان داده تطبیقی-2020
Data stream classification techniques have been playing an important role in big data analytics recently due to their diverse applications (e.g. fraud and intrusion detection, forecasting and healthcare monitoring systems) and the growing number of real-world data stream generators (e.g. IoT devices and sensors, websites and social network feeds). Streaming data is often prone to evolution over time. In this context, the main challenge for computational models is to adapt to changes, known as concept drifts, using data mining and optimisation techniques. We present a novel ensemble technique called RED-PSO that seamlessly adapts to different concept drifts in non-stationary data stream classification tasks. RED-PSO is based on a three-layer architecture to produce classification types of different size, each created by randomly selecting a certain percentage of features from a pool of features of the target data stream. An evolutionary algorithm, namely, Replicator Dynamics (RD), is used to seamlessly adapt to different concept drifts; it allows good performing types to grow and poor performing ones to shrink in size. In addition, the selected feature combinations in all classification types are optimised using a non-canonical version of the Particle Swarm Optimisation (PSO) technique for each layer individually. PSO allows the types in each layer to go towards local (within the same type) and global (in all types) optimums with a specified velocity. A set of experiments are conducted to compare the performance of the proposed method to state-of-the-art algorithms using real-world and synthetic data streams in immediate and delayed prequential evaluation settings. The results show a favourable performance of our method in different environments.
Keywords: Ensemble learning | Data stream mining | Concept drifts | Bio-inspired algorithms | Non-stationary environments | Particle swarm optimisation | Replicator dynamics
HealthFog: An ensemble deep learning based Smart Healthcare System for Automatic Diagnosis of Heart Diseases in integrated IoT and fog computing environments
HealthFog: یک سیستم هوشمند درمانی هوشمند مبتنی بر یادگیری عمیق برای تشخیص خودکار بیماری های قلبی در محیط های IoT و محاسبات مه-2020
Cloud computing provides resources over the Internet and allows a plethora of applications to be deployed to provide services for different industries. The major bottleneck being faced currently in these cloud frameworks is their limited scalability and hence inability to cater to the requirements of centralized Internet of Things (IoT) based compute environments. The main reason for this is that latency-sensitive applications like health monitoring and surveillance systems now require computation over large amounts of data (Big Data) transferred to centralized database and from database to cloud data centers which leads to drop in performance of such systems. The new paradigms of fog and edge computing provide innovative solutions by bringing resources closer to the user and provide low latency and energy efficient solutions for data processing compared to cloud domains. Still, the current fog models have many limitations and focus from a limited perspective on either accuracy of results or reduced response time but not both. We proposed a novel framework called HealthFog for integrating ensemble deep learning in Edge computing devices and deployed it for a real-life application of automatic Heart Disease analysis. HealthFog delivers healthcare as a fog service using IoT devices and efficiently manages the data of heart patients, which comes as user requests. Fog-enabled cloud framework, FogBus is used to deploy and test the performance of the proposed model in terms of power consumption, network bandwidth, latency, jitter, accuracy and execution time. HealthFog is configurable to various operation modes which provide the best Quality of Service or prediction accuracy, as required, in diverse fog computation scenarios and for different user requirements.
Keywords: Fog computing | Internet of things | Healthcare | Deep learning | Ensemble learning | Heart patient analysis
An ensemble learning approach to lip-based biometric verification, with a dynamic selection of classifiers
یک رویکرد یادگیری گروه برای تأیید بیومتریک مبتنی بر لب ، با انتخاب پویای طبقه بندی کننده ها-2019
Machine learning approaches are largely focused on pattern or object classification, where a combination of several classifier systems can be integrated to help generate an optimal or suboptimal classification decision. Multiple classification systems have been extensively developed because a committee of clas- sifiers, also known as an ensemble, can outperform the ensemble’s individual members. In this paper, a classification method based on an ensemble of binary classifiers is proposed. Our strategy consists of two phases: (1) the competence of the base heterogeneous classifiers in a pool is determined, and (2) an ensemble is formed by combining those base classifiers with the greatest competences for the given input data. We have shown that the competence of the base classifiers can be successfully calculated even if the number of their learning examples was limited. Such a situation is particularly observed with biomet- ric data. In this paper, we propose a new biometric data structure, the Sim coefficients, along with an efficient data processing technique involving a pool of competent classifiers chosen by dynamic selection.
Keywords: Lip-based biometrics | Dynamic classifiers selection | Pattern recognition | Ensemble classification | Person verification
Automated diagnosis of ear disease using ensemble deep learning with a big otoendoscopy image database
تشخیص خودکار بیماری گوش با استفاده از یادگیری عمیق گروه با یک پایگاه داده بزرگ تصویر otoendoscopy-2019
Background: Ear and mastoid disease can easily be treated by early detection and appropriate medical care. However, short of specialists and relatively lowdiagnostic accuracy calls for a newway of diagnostic strategy, inwhich deep learning may play a significant role. The current study presents a machine learning model to automatically diagnose ear disease using a large database of otoendoscopic images acquired in the clinical environment. Methods: Total 10,544 otoendoscopic images were used to train nine public convolution-based deep neural networks to classify eardrum and external auditory canal features into six categories of ear diseases, covering most ear diseases (Normal, Attic retraction, Tympanic perforation, Otitis externa±myringitis, Tumor). After evaluating several optimization schemes, two best-performingmodelswere selected to compose an ensemble classifier, by combining classification scores of each classifier. Findings: According to accuracy and training time, transfer learning models based on Inception-V3 and ResNet101 were chosen and the ensemble classifier using the two models yielded a significant improvement over each model, the accuracy of which is in average 93·67% for the 5-folds cross-validation. Considering substantial data-size dependency of classifier performance in the transfer learning, evaluated in this study, the high accuracy in the current model is attributable to the large database. Interpretation: The current study is unprecedented in terms of both disease diversity and diagnostic accuracy, which is compatible or even better than an average otolaryngologist. The classifier was trainedwith data in a various acquisition condition,which is suitable for the practical environment. This study shows the usefulness of utilizing a deep learning model in the early detection and treatment of ear disease in the clinical situation. Fund: This research was supported by Brain Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT(NRF-2017M3C7A1049051).
Keywords: Convolutional neural network | Deep learning | Otoendoscopy | Tympanic membrane | Ear disease | Ensemble learning
CWV-BANN-SVM ensemble learning classifier for an accurate diagnosis of breast cancer
طبقه بندی یادگیری گروه CWV-BANN-SVM برای تشخیص دقیق سرطان پستان-2019
This paper presents a new data mining technique for an accurate prediction of breast cancer (BC), which is one of the major mortality causes among women around the globe. The main objective of our study is to expand an automatic expert system (ES) to provide an accurate diagnosis of BC. Both, Support Vector Machines (SVMs) and Artificial Neural Networks (ANNs) were applied to analyze BC data. The wellknown Wisconsin Breast Cancer Dataset (WBCD), available in the UCI repository, was examined in our study. We first tested the SVM algorithm using various values of the C, e and c parameters. As a result of the first experiment, we were able to observe that the adjustment of these regularization parameters can greatly improve the performance of the traditional SVM algorithm applied for BC detection. The highest obtained accuracy at the first step was 99.71%. Then, we performed a new BC detection approach based on two ensemble learning techniques: the confidence-weighted voting method and the boosting ensemble technique. Our model, called CWV-BANNSVM, combines boosting ANNs (BANN) and two SVMs, using optimal parameters selected during the first experiment. The performance of the applied methods was evaluated using several popular metrics, such as specificity, sensitivity, precision, FPR, FNR, F1 score, AUC, Gini and accuracy. The proposed CWV-BANNSVM model was able to improve the performance of the traditional machine learning algorithms applied to BC detection, reaching the accuracy of 100%. To overcome the overfitting issue, we determined and used some appropriate parameter values of polynomial SVM. Our comparison with the existing studies dedicated to BC prediction suggests that the proposed CWV-BANN-SVM model provides one of the best prediction performances overall.
Keywords: Data mining | Machine learning | Ensemble technique | Breast cancer | Support vector machine | Artificial neural network
Deep ensemble learning based probabilistic load forecasting in smart grids
پیش بینی بار احتمالی مبتنی بر یادگیری گروه عمیق در شبکه های هوشمند-2019
With the availability of fine-grained smart meter data, there has been increasing interest in using this information for ecient and reliable energy management. In particular, accurate probabilistic load forecasting for individual consumers is critical in determining the uncertainties in future demand with the goal of improving smart grid reliability. Compared with the aggregate loads, individual load profiles exhibit higher irregularity and volatility and thus less predictable. To address these challenges, a novel deep ensemble learning based probabilistic load forecasting framework is proposed to quantify the load uncertainties of individual customers. This framework employs the profiles of dierent customer groups integrated into the understanding of the task. Specifically, customers are clustered into separate groups based on their profiles and multitask representation learning is employed on these groups simultaneously. This leads to a better feature learning across groups. Case studies conducted on an open access dataset from Ireland demonstrate the eectiveness and superiority of the proposed framework
Keywords: Deep ensemble learning | multitask representation learning | probabilistic load forecasting | smart grid | customer profiles
Advancing Ensemble Learning Performance through data transformation and classifiers fusion in granular computing context
پیشبرد عملکرد یادگیری گروه از طریق تبدیل داده ها و ترکیب طبقه بندیگرها در زمینه محاسبات دانه ای-2019
Classification is a special type of machine learning tasks, which is essentially achieved by training a clas- sifier that can be used to classify new instances. In order to train a high performance classifier, it is crucial to extract representative features from raw data, such as text and images. In reality, instances could be highly diverse even if they belong to the same class, which indicates different instances of the same class could represent very different characteristics. For example, in a facial expression recognition task, some instances may be better described by Histogram of Oriented Gradients features, while others may be better presented by Local Binary Patterns features. From this point of view, it is necessary to adopt ensemble learning to train different classifiers on different feature sets and to fuse these classi- fiers towards more accurate classification of each instance. On the other hand, different algorithms are likely to show different suitability for training classifiers on different feature sets. It shows again the ne- cessity to adopt ensemble learning towards advances in the classification performance. Furthermore, a multi-class classification task would become increasingly more complex when the number of classes is increased, i.e. it would lead to the increased difficulty in terms of discriminating different classes. In this paper, we propose an ensemble learning framework that involves transforming a multi-class classification task into a number of binary classification tasks and fusion of classifiers trained on different f eature sets by using different learning algorithms. We report experimental studies on a UCI data set on Sonar and the CK + data set on facial expression recognition. The results show that our proposed ensemble learning approach leads to considerable advances in classification performance, in comparison with popular learn- ing approaches including decision tree ensembles and deep neural networks. In practice, the proposed approach can be used effectively to build an ensemble of ensembles acting as a group of expert systems, which show the capability to achieve more stable performance of pattern recognition, in comparison with building a single classifier that acts as a single expert system.
Keywords: Machine learning | Ensemble learning | Classification | Bagging | Boosting | Random forests
Crypto-ransomware early detection model using novel incremental bagging with enhanced semi-random subspace selection
مدل تشخیص زودهنگام رمزنگاری شده با استفاده از کیف های افزایشی جدید با انتخاب زیرزمینی نیمه تصادفی پیشرفته-2019
The irreversible effect is what characterizes crypto-ransomware and distinguishes it from traditional malware. That is, even after neutralizing the attack, the targeted files remain encrypted and cannot be accessed without the decryption key. Thus, it is imperative to detect such a threat early, i.e. in the initial phases before the encryption takes place. However, the lack of sufficient information in initial phases of the attack is the main challenge to early detection, leading to low detection accuracy and a high rate of false alarms. This is due to the way that the existing solutions have been designed based on, which assumes the availability of complete information about the behavior of such attacks at detection time. Nevertheless, this does not hold for early detection that takes place while the attack is underway, and data are not fully available. To address such limitations, this paper proposes two novel techniques; incremental bagging (iBagging) and enhanced semi-random subspace selection (ESRS), and incorporates them into an ensemble-based detection model. The proposed iBagging was firstly used to build incremental subsets in a way that reflects the progression of crypto-ransomware behavior during its different attack phases. ESRS was then used to build optimal, noise-free and diverse features subspaces, by which, a pool of classifiers was trained. Finally, a grid search was employed to select the best combination of base classifiers. Majority voting was utilized for the final decision. The experimental evaluation of the proposed techniques and model was conducted and compared with the existing crypto-ransomware early detection solutions. The results demonstrate that the proposed techniques and model overcame the data limitation in the early phases of the attacks and achieved higher detection accuracy than existing solutions.
Keywords: Crypto-ransomware | Malware | Ransomware | Bitcoin | Cryptography | Early detection | Ensemble learning | IoT
سیستم تشخیص نفوذ توزیع شده برای محیط های ابری بر اساس تکنیک های داده کاوی
سال انتشار: 2018 - تعداد صفحات فایل pdf انگلیسی: 7 - تعداد صفحات فایل doc فارسی: 16
تقریبا دو دهه بعد از ظهور انها؛ محاسبات ابری همچنان در میان سازمان ها و کاربران فردی در حال افزایش است. بسیاری از مسائل امنیتی همراه انتقال برای این الگوی محاسباتی شامل تشخیص نفوذ به وجود می اید. ابزارهای حمله و نفوذ با شکستن سیستم های تشخیص نفوذ سنتی (IDS) با مقدار زیادی از اطلاعات ترافیک شبکه و رفتارهای پویا پیچیده تر شده است. IDSs ابری موجود از کمبود دقت تشخیص؛ نرخ مثبت کاذب بالا و زمان اجرای بالا رنج می برد. در این مقاله ما یک یادگیری توزیع ماشینی بر مبنی سیستم تشخیص نفوذ برای محیط های ابری را ارائه می دهیم. سیستم پیشنهاد شده برای مندرجات در سمت ابری به وسیله اندازه همراه اجزای شبکه لبه از ابرهای ارائه شده است. اینها به ترافیک رهگیری شبکه های ورودی به لبه شبکه routers از از لایه فیزیکی اجازه می دهد. یک الگوریتم پنجره کشویی (sliding window) مبتنی بر زمان برای پیش پردازش شبکه گرفتار ترافیک در هر router ابری استفاده می شود و سپس در نمونه تشخیص ناهنجاری دسته بندی Naive Bayes استفاده می شود. یک مجموعه از گره های سرور کالا بر مبنی یک Hadoop و MapReduce برای هر نمونه تشخیص ناهنجاری از زمانی که تراکم شبکه افزایش می یابد؛ در دسترس است. برای هر پنجره زمانی؛ داده ترافیک ناهنجاری شبکه در هر طرف router برای یک سرور ذخیره سازی مرکزی هماهنگ شده است. بعد؛ یک طبقه بندی یادگیری گروهی بر مبنی یک Forest تصادفی برای اجرای یک مرحله دسته بندی چند کلاسه نهایی به منظور تشخیص انواعی از هر حمله استفاده می شود.
لغات کلیدی: سیستم های تشخیص نفوذ | محاسبات ابری | یادگیری ماشین | هادوپ | MapReduce
|مقاله ترجمه شده|