An incremental intrusion detection system using a new semi‐supervised stream classification method

In recent years, the utilization of machine learning and data mining techniques for intrusion detection has received great attention by both security research communities and intrusion detection system (IDS) developers. In intrusion detection, the most important constraints are the imbalanced class distribution, the scarcity of the labeled data, and the massive amounts of network flows. Moreover, because of the dynamic nature of the network flows, applying static learned models degrades the detection performance significantly over time. In this article, we propose a new semi‐supervised stream classification method for intrusion detection, which is capable of incremental updating using limited labeled data. The proposed method, called the incremental semi‐supervised flow network‐based IDS (ISF‐NIDS), relies on an incremental mixed‐data clustering, a new supervised cluster adjustment method, and an instance‐based learning. The ISF‐NIDS operates in real time and learns new intrusions quickly using limited storage and processing power. The experimental results on the KDD99, Moore, and Sperotto benchmark datasets indicate the superiority of the proposed method compared with the existing state‐of‐the‐art incremental IDSs.     

Asymmetric Semi‐Supervised Boosting Scheme for Interactive Image Retrieval

Support vector machine (SVM) active learning plays a key role in the interactive content‐based image retrieval (CBIR) community. However, the regular SVM active learning is challenged by what we call “the small example problem” and “the asymmetric distribution problem.” This paper attempts to integrate the merits of semi‐supervised learning, ensemble learning, and active learning into the interactive CBIR. Concretely, unlabeled images are exploited to facilitate boosting by helping augment the diversity among base SVM classifiers, and then the learned ensemble model is used to identify the most informative images for active learning. In particular, a bias‐weighting mechanism is developed to guide the ensemble model to pay more attention on positive images than negative images. Experiments on 5000 Corel images show that the proposed method yields better retrieval performance by an amount of 0.16 in mean average precision compared to regular SVM active learning, which is more effective than some existing improved variants of SVM active learning.     

On the expected longest length probe sequence for hashing with separate chaining

Hashing is a widely used technique for data organization. Hash tables enable a fast search of the stored data and are used in a variety of applications ranging from software to network equipment and computer hardware. One of the main issues associated with hashing are collisions that cause an increase in the search time. A number of alternatives have been proposed to deal with collisions. One of them is separate chaining, in which for each hash value an independent list of the elements that have that value is stored. In this scenario, the worst case search time is given by the maximum length of that list across all hash values. This worst case is often referred to as Longest Length Probe Sequence (llps) in the literature. Approximations for the expected longest length probe sequence when the hash table is large have been proposed and an exact analytical solution has also been presented in terms of a set of recurring equations. In this paper, a novel analytical formulation of the expected longest length probe sequence is introduced. The new formulation does not require a recursive computation and can be easily implemented in a symbolic computation tool.     

基于复杂网络上监督学习方法的研究综述

复杂网络已经成为复杂系统分析问题的通用方法,随着人工智能和机器学习的广泛兴起,越来越多的学者开始关注在复杂网络上进行机器学习。监督学习作为机器学习的一个重要组成部分,本文深入研究和总结了基于复杂网络的监督学习方法。首先,本文分别从复杂网络和监督学习的理论基础入手,明确了相似性函数和相异性函数的概念和测度方法,系统梳理了复杂网络的构建方法,并阐明了监督学习的概念及其在机器学习中的地位。其次,介绍了监督学习的几种常用算法,梳理了各种算法的研究现状。然后,提出了基于复杂网络监督学习方法未来关注方向。最后,说明了基于复杂网络监督学习方法的局限性,为相关学者的研究提供了参考。     

Chiral-Quantum-Optics-Based Supervised Learning

Based on the excellent control of single photons realized by atom-photon-chiral couplings,a novel quantum-optics scheme for supervised learning is proposed. The single-photon rotating and phase-shift operations, which can be controlled by another single photon, are realized by proper atom-photon-chiral couplings. Then, an algorithm to perform the supervised learning tasks, composed by integrating the realized gates, and implemented by the tunable gate parameters, is realized.     

基于改进的监督LLE人脸识别算法

LLE是一种无监督的非线性降维方法,广泛应用于人脸特征提取,但是该方法缺乏样本点的类别信息.提出了一种新方法,在LLE的基础上引入有监督的学习机制和增加样本点的类别信息,通过减少类内距离而增加类间距离和最小化局部数据的全局重构误差,同时结合核邻域保持投影方法(KNPP)来提取高维人脸数据的非线性特征.算法有利于分类识别...     

Incremental POP Learning

In recently proposed partial oblique projection (POP) learning, a function space is decomposed into two complementary subspaces, so that functions belonging to one of which can be optimally estimated. This paper shows that when the decomposition is specially performed so that the above subspace becomes the largest, a special learning called SPOP learning is obtained and correspondingly an incremental learning is implemented, result of which equals exactly to that of batch learning including novel data. The effectiveness of the method is illustrated by experimental results.     

Partial Oblique Projection Learning for Optimal Generalization

In some viewpoints, supervised learning is discussed in the framework of function approximation, which means that different criteria result in learning methods of different abilities in generalization [1]. From the standpoint of the original space to which the desired function belongs, projection-based criterion aims directly at the generalization ability[2]. The projection concept is applied to projection learning (PL), partial projection learning (PTPL), and averaged projection learning (A…