Learning Structure Models with Context Information for Visual Tracking

Tracking objects that undergo abrupt appearance changes and heavy occlusions is a challenging problem which conventional tracking methods can barely handle.To address the problem, we propose an online structure learning algorithm that contains three layers: an object is represented by a mixture of online structure models (OSMs) which are learnt from block-based online random forest classifiers (BORFs).BORFs are able to handle occlusion problems since they model local appearances of the target.To further improve the tracking accuracy and reliability, the algorithm utilizes mixture relational models (MRMs) as multi-mode context information to integrate BORFs into OSMs.Furthermore, the mixture construction of OSMs can avoid over-fitting effectively and is more flexible to describe targets.Fusing BORFs with MRMs, OSMs capture the discriminative parts of the target, which guarantees the reliability and robustness of our tracker.In addition, OSMs incorporate with block occlusion reasoning to update our BORFs and MRMs, which can deal with appearance changes and drifting problems effectively.Experiments on challenging videos show that the proposed tracker performs better than several state-of-the-art algorithms.     

An efficient frontier for sum deviation JIT sequencing problem in mixed-model systems via apportionment

In this paper, the sum deviation just-in-time (JIT) sequencing problem in mixed-model production systems is studied relating with the discrete apportionment problem together with their respective mathematical formulations. The assignment formulation for the first problem is brieffy discussed followed by the existence of JIT cyclic sequences. Presenting the concise discussion on divisor methods for the discrete apportionment problem, we have proposed two mean-based divisor functions for this problem claiming that they are better than the existing divisors; hence, we found a better bound for the JIT sequencing problem. The linkage of both the problems is characterized in terms of similar type of objective functions. The problems are shown equivalent via suitable transformations and similar properties. The joint approaches for the two problems are discussed in terms of global and local deviations proposing equitably efficient solution.     

Complete Boolean Satisfiability Solving Algorithms Based on Local Search

Boolean satisfiability (SAT) is a well-known problem in computer science, artificial intelligence, and operations research. This paper focuses on the satisfiability problem of Model RB structure that is similar to graph coloring problems and others. We propose a translation method and three effective complete SAT solving algorithms based on the characterization of Model RB structure. We translate clauses into a graph with exclusive sets and relative sets. In order to reduce search depth, we determine search order using vertex weights and clique in the graph. The results show that our algorithms are much more effective than the best SAT solvers in numerous Model RB benchmarks, especially in those large benchmark instances.     

Deep Semantic Structure of Natural Language

This paper presents the result of research of deep structure of natural language. The main result attained is the existence of a deterministic mathematical model that relates phonetics to associated mental images starting from the simplest linguistic units in agreement with the human response to different acoustic stimuli. Moreover, there exists two level hierarchy for natural language understanding. The first level uncovers the conceptual meaning of linguistic units, and hence forming a corresponding mental image. At the second level the operational meaning is found to suit, context, pragmatics, and world knowledge. This agrees with our knowledge about human cognition. The resulting model is parallel, hierarchical but still concise to explain the speed of natural language understanding.     

The complexity of checking consistency of pedigree information and related problems

Consistency checking is a fundamental computational problem in genetics. Given a pedigree and information on the genotypes (of some) of the individuals in it, the aim of consistency checking is to determine whether these data are consistent with the classic Mendelian laws of inheritance. This problem arose originally from the geneticists‘ need to filter their input data from erroneous information, and is well motivated from both a biological and a sociological viewpoint. This paper shows that consistency checking is NP-complete, even with focus on a single gene and in the presence of three alleles. Several other results on the computational complexity of problems from genetics that are related to consistency checking are also offered. In particular, it is shown that checking the consistency of pedigrees over two alleles, and of pedigrees without loops, can be done in polynomial time.     

Developing a multi-objective,multi-item inventory model and three algorithms for its solution

We develop a multi-objective model in a multi-product inventory system.The proposed model is a joint replenishment problem(JRP) that has two objective functions.The first one is minimization of total ordering and inventory holding costs,which is the same objective function as the classic JRP.To increase the applicability of the proposed model,we suppose that transportation cost is independent of time,is not a part of holding cost,and is calculated based on the maximum of stored inventory,as is the case in many real inventory problems.Thus,the second objective function is minimization of total transportation cost.To solve this problem three efficient algorithms are proposed.First,the RAND algorithm,called the best heuristic algorithm for solving the JRP,is modified to be applicable for the proposed problem.A multi-objective genetic algorithm(MOGA) is developed as the second algorithm to solve the problem.Finally,the model is solved by a new algorithm that is a combination of the RAND algorithm and MOGA.The performances of these algorithms are then compared with those of the previous approaches and with each other,and the findings imply their ability in finding Pareto optimal solutions to 3200 randomly produced problems.     

Constrained optimal steady-state control for isolated traffic intersections

The steady-state or cyclic control problem for a simplified isolated traffic intersection is considered. The optimization problem for the green-red switching sequence is formulated with the help of a discrete-event max-plus model. Two steady-state control problems are formulated： optimal steady-state with green duration constraints, and optimal steady-state control with lost time. In the case when the criterion is a strictly increasing, linear function of the queue lengths, the steady-state control problems can be solved analytically. The structure of constrained optimal steady-state traffic control is revealed, and the effect of the lost time on the optimal solution is illustrated.     

MNP: A Class of NP Optimization Problems

A large class of NP optimization problems called MNP are studied.It is shown that Rmax(2)is in this class and some problems which are not likely in Rmax(2) are in this class.A new kind of reductions,SL-reductions,is defined to preserve approximability and nonapproximability,so it is a more general version of L-reductions and A-reductions.Then some complete problems of this class under SL-reductions are shown and it is proved that the max-clique problem is one of them.So all complete problems in this class are as difficult to approximate as the max-clique problem.     

Semi-online scheduling with machine cost

For most scheduling problems the set of machines is fixed initially and remains unchanged for the duration of the problem.Recently Imreh and Nogaproposed to add the concept of machine cost to scheduling problems and considered the so-called List Model problem.An online algorthm with a competitive ratio 1.618 was given while the lower boud is 4/3.In this paper,two different semi-onlne versions of this problem are studied‘.In the first case,it is assumed that the processing time of the largest job is known a priori.A semi-online algorithm is presented with the competitive ratio at most 1.5309 while the lower bound is 4/3,In the second case,it is assumed that the total processing time of all jobs is known in advance.A semi-online algorithm is presented with the competitive ratio at most 1.414 while the lower bound is 1.161.It is shown that the additional partial available information about the jobs leads to the possibility of constructing a schedule with a smaller competitive ratio than that of online algorithms.     

基于知觉加工模式的发展式分词算法

分词是自然语言理解的一个重要过程,由于语言问题又是最重要的心理学问题之一,所以从认知心理学和发展心理学的观点出发,将分词问题看成是知觉问题在语言信息处理过程中的一部分,把知觉表达和知觉的双向加工过程引入对分词问题的分析,并同时考虑言语发展的心理过程,把对言语发展规律的归纳作为构造学习算法的基础和模板,以进化的方式来逐步改进自然语言理解系统的性能,从而避免知识获取瓶颈在自然语言理解问题中的出现。     

Biocomputing: an insight from linguistics

This paper is placed in a formal framework in which the interdisciplinary study of natural language is conducted by integrating linguistics, computer science and biology. It provides an overview of the field of research, conveying the main biological ideas that have influenced research in linguistics. Our work highlights the main methods of molecular computing that have been applied to the processing and study of the structure of natural language: DNA computing, membrane computing and networks of evolutionary processors. Moreover, some new challenges and lines of research for the future are pointed out, that can provide important improvements in the understanding of natural language as a structure and a human capacity.     

Midwinters, end games, and body parts: a classification of part-whole relations

This paper deals with the conceptual part-whole relation as it occurs in language processing, visual perception, and general problem solving. One important long-term goal is to develop a naive or common sense theory of the mereological domain, that is the domain of parts and wholes and their relations. In this paper, we work towards such a theory by presenting a classification of part-whole relations that is suitable for different cognitive tasks and give proposals for the representation and processing of these relations. In order to be independent of specific tasks like language understanding or the recognition of objects, we use structural properties to develop our classification.The paper starts with a brief overview of the mereological research in different disciplines and two examples of the role of part-whole relations in linguistics (possessive constructions) and knowledge processing (reasoning about objects). In the second section, we discuss two important approaches to mereological problems: the "Classical Extensional Mereology" as presented by Simons and the meronymic system of part-whole relations proposed by Winston, Chaffin and Hermann. Our own work is described in the third and last section. First, we discuss different kinds of wholes according to their inherent compositional structure; complexes, collections, and masses. Then partitions induced by or independent of the compositional structure of a whole are described, accompanied by proposals for their processing.     

Survey on frontiers of language and robotics

ABSTRACT

The understanding and acquisition of a language in a real-world environment is an important task for future robotics services. Natural language processing and cognitive robotics have both been focusing on the problem for decades using machine learning. However, many problems remain unsolved despite significant progress in machine learning (such as deep learning and probabilistic generative models) during the past decade. The remaining problems have not been systematically surveyed and organized, as most of them are highly interdisciplinary challenges for language and robotics. This study conducts a survey on the frontier of the intersection of the research fields of language and robotics, ranging from logic probabilistic programming to designing a competition to evaluate language understanding systems. We focus on cognitive developmental robots that can learn a language from interaction with their environment and unsupervised learning methods that enable robots to learn a language without hand-crafted training data.     

基于词缀的维吾尔谚语识别关键技术研究

在自然语言理解、机器翻译、舆情分析等自然语言处理领域中,维吾尔谚语识别是整个文本实体识别的重要组成部分。为满足维吾尔谚语信息化的需求,本文构建了比较完善的维吾尔谚语语料库。同时,从传统语言学角度对维吾尔谚语的语法、语义结构进行分析,构建了一个由维吾尔谚语功能语类（词缀）组成的、专属维吾尔谚语规则的知识库,并将此知识库与自然语言处理技术相结合,实现一个既能够从文本中识别出维吾尔谚语,又能提供维汉互译等功能的信息软件系统。该系统也为开展计算机理解与处理维吾尔文字奠定了一个崭新的基础。     

基于区分词的汉语隐喻短语识别

隐喻识别是自然语言处理的一个重要研究分支。目前人们越来越清楚地认识到隐喻在思维及语言中所处的中心地位。从计算语言学和自然语言处理的角度来考虑,隐喻问题若不能得到很好的处理,语言理解和机器翻译的效果都会受到影响。通过观察隐喻短语和非隐喻短语在汉语中的上下文发现,有一批词可用于有效地识别隐喻短语,称之为区分词。首先从Web中自动抽取了一部分区分词,进而提出了一种基于区分词的隐喻短语识别方法。实验表明基于区分词的识别方法是有效的。     

汉语结构优先关系的自动获取*

提出了一种用于定量描述歧义结构分布特点的知识表示机制:结构优先关系SPR(structure preference relation),介绍了针对不同语料文本的两种SPR获取方法:树库语料的自动发现和原始文本的自动获取,并且通过不同的实验证明了这些方法的可行性和实用性.另外,还介绍了SPR的若干应用前景,并提出进一步的研究方向.     

基于参照的对词结构操作语义的归纳学习

心理语言学的研究和认知发展过程证明在语言获得的早期经历了一个自主的归纳学习过程,本文的出发点是语言发展的规律,并将词结构形式语义的获得过程和表示基础放在一个具有统一的语言理解和语言产生机制的语言信息加工模型中来考虑。本文讨论了一个基于实例的机器学习系统,为了获得词结构的形式语义,采用了操作语义的定义,并设计了一个基于参照的发现学习算法,其目的是使语义能伴随例句样本的丰富而精密化。     

蒙古语动词计算机生成研究

该文遵循笔者提出的“整词中枢论”思想，基于自然语言复杂特征合一运算理论，方法和技术，通过揭蒙古语动词构成数学结构，提出了蒙古语动词构成算法的数学模型（机理模型），分析了确立规则的各种复杂问题，建立了一个生成蒙古词动词的FSTN，解决了蒙古词动词的计算机生成过程中的语言学“共性”与“个性”恰当结合的问题，实现了蒙古语动词计算机生成精确度，时间度和空间度三要素的有机统一。