دانلود رایگان مقاله لاتین طبقه بندی نیمه نظارتی افزایشی جریان داده از سایت الزویر
عنوان فارسی مقاله:
طبقه بندی نیمه نظارتی افزایشی جریان های داده ها از طریق خود عامل انتخابی
عنوان انگلیسی مقاله:
Incremental Semi-Supervised classification of data streams via self-representative selection
سال انتشار : 2016
برای دانلود رایگان مقاله طبقه بندی نیمه نظارتی افزایشی جریان داده اینجا کلیک نمایید.
مقدمه انگلیسی مقاله:
1. Introduction
Data today is more deeply woven into the fabric of our daily lives than ever before due to the rapid improvement of digital technology of storage and information processing. Very recent few years have witnessed an explosive growth of data, where continuously collected data streams accounts for a large and important part[1,2]. From the perspectives of computation and machine intelligence, one should establish a data-driven machine that is capable of incrementally analyzing large-scale dynamic data stream, and accumulating knowledge incrementally over time to benefit future learning and decision-making process [3–11]. Consequently, a machine learning paradigm, Incremental Learning (InLe), is developed where the learning process takes place according to the newly emerged examples [12–21]. Compared with traditional supervised learning, InLe is capable of learning new information from sequential examples to facility the decision-making process. It is very suitable for applications where examples do not always arrive simultaneously, and the newly arriving data may bring a new perspective, may even change the statistical distribution of data. Moreover, from the biological viewpoint, InLe is more consistent with human learning where human beings already use possessed knowledge along with the experiences for learning and decision making.Nowadays many incremental learning architectures [22,23] and algorithms [12–15,20,21,35] have been developed to deal with data streams, which can be categorized as Absolute Incremental Learning (AInLe) and Selective Incremental Learning (SInLe). In AInLe, new data are analyzed separately, and new features are formed and combined with the existing ones. In SInLe, the selected training set based on the proximity and impact of new data and new information are retrained in light of new information. Most of available InLe approaches are SInLe, which do not assume the availability of a sufficient labeled dataset before the learning, but the training examples appear over time. However, in real-life scenarios, new examples are not always labeled timely. In practical, massive amounts of data are collected dynamically in very rapid mode, resulting in the difficulty of offering labeled samples over time. For example, labeling examples from surveillance and mobile sensor network data streams is infeasible both in time and resource. On the other hand, preparing a sufficiently large number of labeled training samples at the very beginning is practically impossible, for the changing environment where new characteristic of samples or even new kind of samples are generated over time. Consequently, it is necessary to automatically update an existing training set in an incremental fashion to accommodate new information, by adding newly emerged samples to the training set.
برای دانلود رایگان مقاله طبقه بندی نیمه نظارتی افزایشی جریان داده اینجا کلیک نمایید.
کلمات کلیدی:
[PDF]Semi-supervised Learning Using Incremental Support Vector ... - UKSim uksim.info/uksim2016/CD/data/0888a184.pdf by H Al-Behadili - Cited by 1 - Related articles proposed approach. Keywords – data streaming; online classification; incremental ... Semi- supervised learning denotes training of the classifier using ... self-training [1]. Incremental ..... in three axes, a code representing the direction of the. An incremental intrusion detection system using a new semi ... onlinelibrary.wiley.com/doi/10.1002/dac.3002/full Jun 19, 2015 - The network flows are stream based; thus, new attack types may be emerging. ... With respect to incremental IDS, semi-supervised learning should apply limited ... mixed-data clustering algorithm, called the mixed self-organizing ... are the representative labeled instances that are applied to classify new ... A semi-supervised weighted clustering framework facing to hybrid ... ieeexplore.ieee.org/document/5554528/ by X Chen - 2010 - Cited by 1 - Related articles ... clustering analysis problem of infinite hybrid attributes data streams in finite space ... on decision-tree classify of little sample using a semi-supervised strategy. Unsupervised Classification of Data Streams based on Typicality and ... eprints.lancs.ac.uk/81699/1/bare_conf.pdf by BSJ Costa - 2016 - Related articles a self-evolving fuzzy-rule-basis. ... data. It can perform fuzzy classification/soft-labeling, which is preferred over ... supervised/semi-supervised/unsupervised classification based .... represented by a fuzzy set, in AnYa systems, the antecedent. Liying Jiao - Publications - The Academic Family Tree https://academictree.org/chemistry/publications.php?pid=485532 2016, Liu H, Zhu D, Yang S, Hou B, Gou S, Xiong T, Jiao L. Semisupervised Feature .... Semi-Supervised classification of data streams via self-representative ...
