快速深度学习的鲁棒视觉跟踪

目的 基于深度学习的视觉跟踪算法具有跟踪精度高、适应性强的特点,但是,由于其模型参数多、调参复杂,使得算法的时间复杂度过高。为了提升算法的效率,通过构建新的网络结构、降低模型冗余,提出一种快速深度学习的算法。方法 鲁棒特征的提取是视觉跟踪成功的关键。基于深度学习理论,利用海量数据离线训练深度神经网络,分层提取描述图像的特征;针对网络训练时间复杂度高的问题,通过缩小网络规模得以大幅缓解,实现了在GPU驱动下的快速深度学习;在粒子滤波框架下,结合基于支持向量机的打分器的设计,完成对目标的在线跟踪。结果 该方法精简了特征提取网络的结构,降低了模型复杂度,与其他基于深度学习的算法相比,具有较高的时效性。系统的跟踪帧率总体保持在22帧/s左右。结论 实验结果表明,在目标发生平移、旋转和尺度变化,或存在光照、遮挡和复杂背景干扰时,本文算法能够实现比较稳定和相对快速的目标跟踪。但是,对目标的快速移动和运动模糊的鲁棒性不够高,容易受到相似物体的干扰。     

Strong separation of learning classes

Abstract

Suppose LC ₁ and LC ₂ are two machine learning classes each based on a criterion of success. Suppose, for every machine which learns a class of functions according to the LC ₂ criterion of success, there is a machine which learns this class according to the LC ₂ criterion. In the case where the converse does not hold LC, is said to be separated from LC ₂. It is shown that for many such separated learning classes from the literature a much stronger separation holds: (?𝒞∈LC ₁) (?𝒞' ∈LC ₂ - LC ₁(( [' ?𝒞] It is also shown that there is a pair of separated learning classes from the literature for which the stronger separation above does not hold. A philosophical heuristic toward the design of artificially intelligent learning programs is presented with each strong separation result.     

Adaptive behaviorin cellular automata using rough set theory

The paper uses ideas from machine learning and artificial intelligence to provide the model of cellular automata based on rough set theory and the response to it in simulated cars. Recently, the examination and modeling of vehicular traffic has become an important subject of research. We propose in this paper, a road-traffic system based on two-dimensional cellular automata combined with rough set theory, to model the flow and jamming that is common in an urban environment. The modeled development process in this paper involves simulated processes of evolution, learning, and self-organization. The main value of the model is that it provides an illustration of how simple learning processes may lead to the formation of the state machine behavior, which can give emergence to the model.     

Theoretical analyses of cross-validation error and voting in instance-based learning

Abstract

This paper begins with a general theory of error in cross-validation testing of algorithms for supervised learning from examples. It is assumed that the examples are described by attribute-value pairs, where the values are symbolic. Cross-validation requires a set of training examples and a set of testing examples. The value of the attribute that is to be predicted is known to the learner in the training set, but unknown in the testing set. The theory demonstrates that cross-validation error has two components: error on the training set (inaccuracy) and sensitivity to noise (instability). This general theory is then applied to voting in instance-based learning. Given an example in the testing set, a typical instance-based learning algorithm predicts the designated attribute by voting among the k nearest neighbours (the k most similar examples) to the testing example in the training set. Voting is intended to increase the stability (resistance to noise) of instance-based learning, but a theoretical analysis shows that there are circumstances in which voting can be destabilising. The theory suggests ways to minimize cross-validation error, by insuring that voting is stable and does not adversely affect accuracy.     

A systematic review of machine learning techniques for software fault prediction

BackgroundSoftware fault prediction is the process of developing models that can be used by the software practitioners in the early phases of software development life cycle for detecting faulty constructs such as modules or classes. There are various machine learning techniques used in the past for predicting faults.MethodIn this study we perform a systematic review of studies from January 1991 to October 2013 in the literature that use the machine learning techniques for software fault prediction. We assess the performance capability of the machine learning techniques in existing research for software fault prediction. We also compare the performance of the machine learning techniques with the statistical techniques and other machine learning techniques. Further the strengths and weaknesses of machine learning techniques are summarized.ResultsIn this paper we have identified 64 primary studies and seven categories of the machine learning techniques. The results prove the prediction capability of the machine learning techniques for classifying module/class as fault prone or not fault prone. The models using the machine learning techniques for estimating software fault proneness outperform the traditional statistical models.ConclusionBased on the results obtained from the systematic review, we conclude that the machine learning techniques have the ability for predicting software fault proneness and can be used by software practitioners and researchers. However, the application of the machine learning techniques in software fault prediction is still limited and more number of studies should be carried out in order to obtain well formed and generalizable results. We provide future guidelines to practitioners and researchers based on the results obtained in this work.     

主动学习的多标签图像在线分类

目的在多标签有监督学习框架中,构建具有较强泛化性能的分类器需要大量已标注训练样本,而实际应用中已标注样本少且获取代价十分昂贵。针对多标签图像分类中已标注样本数量不足和分类器再学习效率低的问题,提出一种结合主动学习的多标签图像在线分类算法。方法基于min-max理论,采用查询最具代表性和最具信息量的样本挑选策略主动地选择待标注样本,且基于KKT(Karush-Kuhn-Tucker)条件在线地更新多标签图像分类器。结果在4个公开的数据集上,采用4种多标签分类评价指标对本文算法进行评估。实验结果表明,本文采用的样本挑选方法比随机挑选样本方法和基于间隔的采样方法均占据明显优势;当分类器达到相同或相近的分类准确度时,利用本文的样本挑选策略选择的待标注样本数目要明显少于采用随机挑选样本方法和基于间隔的采样方法所需查询的样本数。结论本文算法一方面可以减少获取已标注样本所需的人工标注代价;另一方面也避免了传统的分类器重新训练时利用所有数据所产生的学习效率低下的问题,达到了当新数据到来时可实时更新分类器的目的。     

Combinatorial probability and the tightness of generalization bounds

Accurate prediction of the generalization ability of a learning algorithm is an important problem in computational learning theory. The classical Vapnik-Chervonenkis (VC) generalization bounds are too general and therefore overestimate the expected error. Recently obtained data-dependent bounds are still overestimated. To find out why the bounds are loose, we reject the uniform convergence principle and apply a purely combinatorial approach that is free of any probabilistic assumptions, makes no approximations, and provides an empirical control of looseness. We introduce new data-dependent complexity measures: a local shatter coefficient and a nonscalar local shatter profile, which can give much tighter bounds than the classical VC shatter coefficient. An experiment on real datasets shows that the effective local measures may take very small values; thus, the effective local VC dimension takes values in [0, 1] and therefore is not related to the dimension of the space. Konstantin Vorontsov. Born 1971. Graduated from the Faculty of Control and Applied Mathematics, Moscow Institute of Physics and Technology, in 1994. Received candidate’s degree in 1999. Currently is with the Dorodnicyn Computing Centre, Russian Academy of Sciences. Deputy director for research of Forecsys company (). Scientific interests: computational learning theory, machine learning, data mining, probability theory, and combinatorics. Author of 40 papers. Homepage:     

Ignoring data may be the only way to learn efficiently

Abstract

In designing learning algorithms it seems quite reasonable to construct them in a way such that all data the algorithm already has obtained are correctly and completely reflected in the hypothesis the algorithm outputs on these data. However, this approach may totally fail, i.e. it may lead to the unsolvability of the learning problem, or it may exclude any efficient solution of it. In particular, we present a natural learning problem and prove that it can be solved in polynomial time if and only if the algorithm is allowed to ignore data.     

Ensembles and locality: Insight on improving software effort estimation

ContextEnsembles of learning machines and locality are considered two important topics for the next research frontier on Software Effort Estimation (SEE).ObjectivesWe aim at (1) evaluating whether existing automated ensembles of learning machines generally improve SEEs given by single learning machines and which of them would be more useful; (2) analysing the adequacy of different locality approaches; and getting insight on (3) how to improve SEE and (4) how to evaluate/choose machine learning (ML) models for SEE.MethodA principled experimental framework is used for the analysis and to provide insights that are not based simply on intuition or speculation. A comprehensive experimental study of several automated ensembles, single learning machines and locality approaches, which present features potentially beneficial for SEE, is performed. Additionally, an analysis of feature selection and regression trees (RTs), and an investigation of two tailored forms of combining ensembles and locality are performed to provide further insight on improving SEE.ResultsBagging ensembles of RTs show to perform well, being highly ranked in terms of performance across different data sets, being frequently among the best approaches for each data set and rarely performing considerably worse than the best approach for any data set. They are recommended over other learning machines should an organisation have no resources to perform experiments to chose a model. Even though RTs have been shown to be more reliable locality approaches, other approaches such as k-Means and k-Nearest Neighbours can also perform well, in particular for more heterogeneous data sets.ConclusionCombining the power of automated ensembles and locality can lead to competitive results in SEE. By analysing such approaches, we provide several insights that can be used by future research in the area.     

“Old” theories, “New” technologies: Understanding knowledge sharing and learning in Brazilian software development companies

ContextNew technologies such as social networks, wikis, blogs and other social software enable collaborative work and are important facilitators of the learning process. They provide a simple mechanism for people to communicate and collaborate and thus support the creation of knowledge. In software-development companies they are used to creating an environment in which communication and collaboration between workers take place more effectively.ObjectiveThis paper identifies the main tools and technologies used by software-development companies in Brazil to manage knowledge and attempts to determine how these tools and technologies relate to important knowledge-sharing and learning theories and how they support the concepts described by these theories.MethodA survey was conducted in a group of Brazilian software development companies with high levels of process software maturity to see how they implement the Brazilian Software Processes Improvement model (MPS.Br) and use new tools and technologies. The survey used a qualitative analysis to identify which tools are used most and how frequently employees use them. The results of the analysis were compared with data from the literature on three knowledge-sharing and learning theories to understand how the use of these tools relates to the concepts proposed in these theories.ResultsThe results show that some of the tools used by the companies do not apply the concepts described in the theories as they do not help promote organizational learning. Furthermore, although the companies have adopted the tools, these are not often used, mainly because they are felt not to organize information efficiently.ConclusionThe use of certain tools can help promote several concepts described in the theories considered. Moreover, the use of these tools can help reduce the impact of, some common organizational problems. However, companies need to improve existing organizational policies that encourage employees to use these tools more regularly.     

Exploring the effects of emotional design and emotional intelligence in multimedia-based learning: an engineering educational perspective

ABSTRACT

Emotions are an important aspect in learning and with the current boom in instructional technology, researchers are exploring methods to investigate how emotions may be manipulated to positively influence online learning. One such method is by adapting the theory of emotional design through multimedia elements. This theory emphasises on individuality and metacognition in exploring these learning outcomes and by this we choose to explore the effects of emotional intelligence (EI). We replicated the methodology used in previous research studies in emotional design in multimedia learning by further exploring the gaps from those studies especially the effects of negative design, EI and a new sample that primarily focusses on engineering undergraduates in Malaysia. This study was designed as a quantitative quasi-experimental using a 3?×?2 factorial design. Based on the findings, it was found that emotional design is a better predictor of cognitive outcomes, whereas EI was a better predictor of emotional outcomes such as motivation and satisfaction for multimedia-based learning. It was also found that positive and negative designs have similar effects on students’ learning outcomes, while EI affected perceived satisfaction in each design.     

ENIGMA AND THE POLISH CONTRIBUTION: BOOK REVIEW

Abstract

In this article, the authors present a mathematical scavenger hunt designed to motivate and excite students learning RSA cryptography in an introductory number theory course. The hunt relies on the RSA cryptosystem, in which Maple is used to encipher and decipher secret information contained within the clues.     

混合生成式和判别式模型的图像自动标注

目的 由于图像检索中存在着低层特征和高层语义之间的“语义鸿沟”,图像自动标注成为当前的关键性问题.为缩减语义鸿沟,提出了一种混合生成式和判别式模型的图像自动标注方法.方法 在生成式学习阶段,采用连续的概率潜在语义分析模型对图像进行建模,可得到相应的模型参数和每幅图像的主题分布.将这个主题分布作为每幅图像的中间表示向量,那么图像自动标注的问题就转化为一个基于多标记学习的分类问题.在判别式学习阶段,使用构造集群分类器链的方法对图像的中间表示向量进行学习,在建立分类器链的同时也集成了标注关键词之间的上下文信息,因而能够取得更高的标注精度和更好的检索效果.结果 在两个基准数据集上进行的实验表明,本文方法在Corel5k数据集上的平均精度、平均召回率分别达到0.28和0.32,在IAPR-TC12数据集上则达到0.29和0.18,其性能优于大多数当前先进的图像自动标注方法.此外,从精度—召回率曲线上看,本文方法也优于几种典型的具有代表性的标注方法.结论 提出了一种基于混合学习策略的图像自动标注方法,集成了生成式模型和判别式模型各自的优点,并在图像语义检索的任务中表现出良好的有效性和鲁棒性.本文方法和技术不仅能应用于图像检索和识别的领域,经过适当的改进之后也能在跨媒体检索和数据挖掘领域发挥重要作用.     

Developing hybrid intelligent tutoring and hypertext systems

Abstract

The objective of this paper is to present an alternative paradigm to the traditional Knowledge Based Expert Systems Paradigm for developing a full-scale Intelligent Tutoring System that has dominated for years Intelligent Tutoring Systems development. This alternative paradigm which integrates Minsky's Frames with hypertext has been successfully deployed so far in the development of PEDRO, an Intelligent Tutoring System for foreign language learning, SONATA, an Intelligent Tutoring System for music theory learning and INTUITION, an Intelligent Tutoring System for Gaming-Simulation.     

近年目标跟踪算法短评——相关滤波与深度学习

目的 目标跟踪是计算机视觉领域的重要组成部分。近年来,基于相关滤波和深度学习的目标跟踪算法层出不穷,本文拟对经典的若干目标跟踪算法进行阐述与分析。方法 首先,对基于相关滤波跟踪算法的基础理论进行介绍,针对相关滤波算法在特征改进类、尺度改进类、消除边界效应类、图像分块类与目标响应自适应类方面进行总结;接下来,从3个方面对基于深度学习的目标跟踪算法进行阐述与分析：目标分类、结构化回归、孪生网络,并对有代表性的跟踪算法的优势与缺陷进行较深层次的解读。结果 通过列举跟踪算法在相关滤波阶段、深度学习阶段针对不同的改进机制的改进算法,总结各阶段算法的优缺点。对目标跟踪算法的最新进展进行阐述,最终对目标跟踪算法的未来发展方向进行总结。结论 基于相关滤波的目标算法在实时性方面表现优秀,但对于复杂背景干扰、相似物遮挡等情况仍然需要优化。深层的卷积特征对于目标有强大的表示力,通过使相关滤波算法与深度学习结合,大幅度提升了算法表现力。基于深度学习的跟踪算法则更侧重于跟踪的性能,大多无法满足实时性。孪生神经网络的使用对于基于深度学习类目标跟踪算法产生了很大的推动,兼顾了算法的性能和实时性。     

Semiotic clarity of Case Management Model and Notation (CMMN)

ContextBusiness processes are an important asset of every organisation and can be divided into a group of structured processes (i.e. procedural) and loosely structured ones (i.e. declarative). A good example of procedural notations is Business Process Model and Notation (BPMN), a technologically accepted Standard, used by many organisations. A representative of declarative models is a fairly new specification, Case Management Model and Notation (CMMN), used for presenting cases.AimsThe main aim of this article is twofold. Firstly, we will provide an overview of the CMMN standard. Secondly, we will determine the degree of acceptance of CMMN with loosely structured processes by considering the acceptance of BPMN for structured processes.MethodThe acceptance of CMMN will be determined by relying on the theory for identifying cognitive effectiveness, which is one of the indicators for technology acceptance. We used the semiotic clarity principle to identify the degree of compliance for CMMN.ResultsThe results show that 24% of all CMMN elements defined in Specifications, Version 1.1, ensure semiotic clarity. The rest of the elements were classified as one of four anomalies, which are also acceptable to some extent and do not create additional confusion or incomprehensibility to CMMN.ConclusionSemiotic clarity and theory of cognitive effectiveness are good indicators for detecting the degree of technology acceptance of a notation. CMMN is a relatively new notation, and our research is offering a brief presentation of CMMN while also providing a detailed analysis of semiotic clarity, which is part of the theory for cognitive effectiveness.     

Investigating the interrelation between attitudes,learning readiness,and learning styles under virtual learning environment: a study among Indian students

ABSTRACT

Over the last few years there has been an increased emphasis on conceptualising, developing, and implementing virtual Learning initiatives across the Indian education sector. Effective and successful use of Virtual Learning Environments (VLEs) in academic setting requires identifying and solving many important social and behavioural issues. Without identifying and addressing these issues, their use can compound the mistakes of the past and leave the learner with a passive, un-engaging experience, leading to incomplete learning and low performance. Therefore, educators are required to recognise learner’s attitude, readiness, and learning style to take advantage of VLE. This study was conducted to find interrelationship of student attitude, readiness, and learning style towards the effective use of VLE. A total of 240 students who are using any form of VLE for their learning and research volunteered for the research, and the scales were filled via the convenience sampling method. The scales of learning readiness, attitude, and learning style were administered to the participants. The relationship among the variables of the research and the research hypotheses for the model were tested using different statistical tools like correlation, t-test and regression. The research result revealed that there is a correlation among learning readiness, attitude, and learning style, and each of these factors influence the effective use of VLE.     

Implementing multiple intelligences theory in a grade 9 science classroom: The experiences of a high school teacher

Abstract

Multiple intelligences (MI) theory (Gardner, 1983), a pluralistic conception of intelligence, has been adopted by educators in a variety of ways to improve teaching and learning. Some adaptations have included exploring learning styles, designing lesson plans, developing integrated curricula, and adopting authentic forms of assessment.

In this case study, a high school science teacher, Jade (a pseudonym), explored MI theory and used it as a conceptual framework to make decisions about how to structure her Grade 9 science curriculum. In studying the case, a variety of data sources and methods were used, including audiotaped planning meetings and interviews, classroom observations, teacher‐generated concept maps, and reflective journals.

This article describes Jade's initial explorations of the theory, how she incorporated the theory into her daily teaching, and the outcomes of the study for Jade and her students. In addition, tensions and challenges experienced by Jade in applying the theory to practice and the implications for science teaching and learning are discussed.