PERFORMANCE IMPROVEMENT USING ADAPTIVE LEARNING ITINERARIES

In this paper, Bayesian‐Networks (BN) and Ant Colony Optimization (ACO) techniques are combined to find the best path through a graph representing all available itineraries to acquire a professional competence. The combination of these methods allows us to design a dynamic learning path, useful in a rapidly changing world. One of the most important advances in this work is that the amount of pheromones released is variable. This amount is calculated by taking into account the results acquired in the last completed course in relation to the minimum score required. By using ACO and BN, a fitness function, responsible of automatically selecting the next course in the learning graph, is defined. This is done by generating a path that maximizes the probability of each user's success in the course. Therefore, the path can change to improve learners’ average performance, taking into account the pedagogical weight of each learning unit and the social behavior of the system. Furthermore, a discrete dynamical system is obtained and its stability is studied. How to wrap an existing Learning Management System is also described in this work. Finally, an experiment compares this approach with the old on‐line learning system being used previously.     

基于蚁群优化算法的碎纸拼接

对于碎纸拼接问题,找到一个全局一致的最终解决方案是非常关键的。本文提出了一种基于蚁群优化算法(ACO)的全局拼接方法。首先运用基于ψ-s分析的局部匹配方法进行局部拼接,然后利用局部拼接产生的候选匹配对构建搜索图、信息素与节点相关联。在迭代过程中,利用候选匹配对之间的矛盾和由蚁群构建的全局拼接路径更新信息素:对于在迄今最优路径上的候选匹配对,人工蚂蚁释放信息素;而位于构建失败的路径上的候选匹配对的信息素以一定比例蒸发。候选匹配对的信息素最终向一定方向收敛。最后,根据信息素来筛选候选匹配对。实验证明了这种方法的有效性。     

BIC-based node order learning for improving Bayesian network structure learning

Node order is one of the most important factors in learning the structure of a Bayesian network (BN) for probabilistic reasoning. To improve the BN structure learning, we propose a node order learning algorithmbased on the frequently used Bayesian information criterion (BIC) score function. The algorithm dramatically reduces the space of node order and makes the results of BN learning more stable and effective. Specifically, we first find the most dependent node for each individual node, prove analytically that the dependencies are undirected, and then construct undirected subgraphs U_G. Secondly, the U_G- is examined and connected into a single undirected graph U_GC. The relation between the subgraph number and the node number is analyzed. Thirdly, we provide the rules of orienting directions for all edges in UGC, which converts it into a directed acyclic graph (DAG). Further, we rank the DAG’s topology order and describe the BIC-based node order learning algorithm. Its complexity analysis shows that the algorithm can be conducted in linear time with respect to the number of samples, and in polynomial time with respect to the number of variables. Finally, experimental results demonstrate significant performance improvement by comparing with other methods.     

Estimating time and score uncertainty in generating successful learning paths under time constraints

This paper addresses the problem of course (path) generation when a learner's available time is not enough to follow the complete course. We propose a method to recommend successful paths regarding a learner's available time and his/her knowledge background. Our recommender is an instance of long term goal recommender systems (LTRS). This method, after locating a target learner in a course graph, applies a depth‐first search algorithm to find all paths for the learner given a time limitation. In addition, our method estimates learning time and score for all paths. It also indicates the probability of error for the estimated time and score for each path. Finally, our method recommends a path that satisfies the learner's time restriction while maximizing expected learning score. In order to evaluate our proposals for time and score estimation, we used the mean absolute error and average MAE. We have evaluated time and score estimation methods, including one proposed in the literature, on two E‐learning datasets.     

New skeleton-based approaches for Bayesian structure learning of Bayesian networks

Automatically learning the graph structure of a single Bayesian network (BN) which accurately represents the underlying multivariate probability distribution of a collection of random variables is a challenging task. But obtaining a Bayesian solution to this problem based on computing the posterior probability of the presence of any edge or any directed path between two variables or any other structural feature is a much more involved problem, since it requires averaging over all the possible graph structures. For the former problem, recent advances have shown that search + score approaches find much more accurate structures if the search is constrained by a previously inferred skeleton (i.e. a relaxed structure with undirected edges which can be inferred using local search based methods). Based on similar ideas, we propose two novel skeleton-based approaches to approximate a Bayesian solution to the BN learning problem: a new stochastic search which tries to find directed acyclic graph (DAG) structures with a non-negligible score; and a new Markov chain Monte Carlo method over the DAG space. These two approaches are based on the same idea. In a first step, both employ a previously given skeleton and build a Bayesian solution constrained by this skeleton. In a second step, using the preliminary solution, they try to obtain a new Bayesian approximation but this time in an unconstrained graph space, which is the final outcome of the methods. As shown in the experimental evaluation, this new approach strongly boosts the performance of these two standard techniques proving that the idea of employing a skeleton to constrain the model space is also a successful strategy for performing Bayesian structure learning of BNs.     

求解TSP的改进信息素二次更新与局部优化蚁群算法

针对蚁群（ACO）算法收敛速度慢、容易陷入局部最优的缺陷,提出了一种改进信息素二次更新局部优化蚁群算法（IPDULACO）。该算法对蚁群搜索到的当前全局最优解中路径贡献度大于给定的路径贡献阈值的子路径信息素进行二次更新,以提高构成潜在最优解的子路径被选择的概率,从而加快算法的收敛。然后,在搜索过程中,当蚁群陷入局部最优时,使用随机插入法对局部最优解中城市的排序进行调整,以增强算法跳出局部最优解的能力。将改进算法应用于若干经典的旅行售货商问题（TSP）进行仿真实验,实验结果表明,对于小规模的TSP,IPDULACO可以在较少的迭代次数内获得已知最优解;对于较大规模的TSP,IPDULACO可以在较少的迭代次数内获得更精确的解。因此,IPDULACO具有更强的搜索全局最优解的能力和更快的收敛速度,可以高效求解TSP。     

Cellular ants: A method to create collision free trajectories for a cooperative robot team

Creating collision-free trajectories for mobile robots, known as the path planning problem, is considered to be one of the basic problems in robotics. In case of multiple robotic systems, the complexity of such systems increases proportionally with the number of robots, due to the fact that all robots must act as one unit to complete one composite task, such as retaining a specific formation. The proposed path planner employs a combination of Cellular Automata (CA) and Ant Colony Optimization (ACO) techniques in order to create collision-free trajectories for every robot of a team while their formation is kept immutable. The method reacts with obstacle distribution changes and therefore can be used in dynamical or unknown environments, without the need of a priori knowledge of the space. The team is divided into subgroups and all the desired pathways are created with the combined use of a CA path planner and an ACO algorithm. In case of lack of pheromones, paths are created using the CA path planner. Compared to other methods, the proposed method can create accurate collision-free paths in real time with low complexity while the implemented system is completely autonomous. A simulation environment was created to test the effectiveness of the applied CA rules and ACO principles. Moreover, the proposed method was implemented in a system using a real world simulation environment, called Webots. The CA and ACO combined algorithm was applied to a team of multiple simulated robots without the interference of a central control. Simulation and experimental results indicate that accurate collision free paths could be created with low complexity, confirming the robustness of the method.     

Evaluation of an adaptive and intelligent educational hypermedia for enhanced individual learning of mathematics: A qualitative study

The purpose of this study is to design a personalized adaptive and intelligent web based tutoring system based on learning style and expert system named UZWEBMAT and to evaluate its effects on 10th grade students’ learning of the unit of probability. In the study, initially, learning objects were prepared in three different ways in relation to three sub-learning areas of Visual–Auditory–Kinesthetic (VAK) learning style for each subject of the probability unit. These were appropriate for secondary school mathematics curricula. Then, they were transferred into the digital environment. Each student’s dominant learning style determines the content to which s/he will be directed since s/he is directed to the content that is appropriate for his/her learning style. The course to be followed by the students within UZWEBMAT and their browsing around the pages are decided by expert system integrated into the system. This expert system sets the situations in which s/he will get solution supports and the course s/he will follow in accordance with the performance of the student. Hereby, each student may follow a different course, and the solution supports s/he will get may also differ highlighting the individual learning. The sample of the study consists of 81 10th grade students and 3 mathematics teachers from two high schools in Trabzon, Turkey. Qualitative data were obtained both from the teachers and students participating in the study in order to answer the research questions about the implementation and evaluation of UZWEBMAT for mathematics teaching in a high school classroom. Obtained data were analyzed using qualitative data analysis methods. According to the results of the present study, positive opinions of students and teachers such as taking into account the individual learning differences and deriving mathematical relations and formulas through exploration became prominent. In addition, there were also other positive opinions of students and teachers such as providing permanent learning and introducing learning responsibility to the students. In this sense, it was concluded that UZWEBMAT is a beneficial instrument for both students and teachers.     

基于自编码器的贝叶斯网嵌入及概率推理

贝叶斯网(BN)是不确定性知识表示和推理的基本框架,广泛用于社交网络、知识图谱和医疗诊断等领域.特定领域中基于BN的分析诊断和决策支持,其核心计算任务是基于BN进行多次概率推理.然而,使用传统的概率推理方法,基于同一BN的多次概率推理其中间过程存在很多重复的计算结果,具有较高的时间复杂度.为了提高多次概率推理的效率,提出易于重用和易于计算的贝叶斯网嵌入及相应的概率推理方法.首先,借鉴图嵌入的基本思想,使用点互信息矩阵来表示BN的有向无环图结构和条件概率参数,提出基于自编码器和注意力机制的BN嵌入方法.其中,自编码器的每一编码层利用节点与其邻居节点(父节点和子节点)的相关性生成节点嵌入,从而在嵌入向量中保存BN节点间的概率依赖关系.然后,使用嵌入向量之间的距离来度量节点之间的联合概率,提出基于嵌入向量的BN概率推理方法.实验证明,针对BN的多次概率推理,所提方法的效率高于现有方法,且能得到准确的推理结果.     

启发式强化学习机制的异构双种群蚁群算法

针对传统蚁群算法在解决TSP问题时易陷入局部最优、收敛速度较慢的问题,提出了一种基于启发式强化学习的异构双种群蚁群算法。蚁群分为主种群和子种群,主种群负责解的构建和信息素的更新,子种群则是在构建解的同时对主种群的解集进行替换。算法初期利用启发式算子自适应地控制两个种群的交流频率,通过偏离度系数控制解的交换方式。前期让子种群的最优解去替换主种群的随机解,增加解的多样性,同时引入强化学习机制对交流后主种群最优路径上的信息素进行自适应的奖赏,以增大最优公共路径以后被选择的概率。后期则控制子种群的最优解去替换主种群的最差解,强化最优路径上信息素的量,并对主种群最优路径上的信息素进行奖赏,进一步提高算法的收敛速度。实验仿真表明,算法能够有效地跳出局部最优,并且解的质量在大规模测试集上有明显的改善。     

动态调整路径选择的蚁群优化算法

针对蚁群算法收敛速度慢和存在停滞现象的缺点,提出对比度增强的路径选择规则以增强其全局搜索能力,选择规则加强了对反馈信息的利用,能加快算法的收敛速度,通过信息熵来动态控制对比度增强的方向,在避免算法停滞的同时加快了算法的收敛速度。将改进后的蚁群优化算法与传统的蚁群优化算法进行比较,仿真实验结果表明,改进算法具有较好的稳定性和全局优化性能,且收敛速度较快。     

Comparative study of pheromone control heuristics in ACO algorithms for solving RCPSP problems

Constraint Satisfaction Problems (CSP) belong to a kind of traditional NP-hard problems with a high impact on both research and industrial domains. The goal of these problems is to find a feasible assignment for a group of variables where a set of imposed restrictions is satisfied. This family of NP-hard problems demands a huge amount of computational resources even for their simplest cases. For this reason, different heuristic methods have been studied so far in order to discover feasible solutions at an affordable complexity level. This paper elaborates on the application of Ant Colony Optimization (ACO) algorithms with a novel CSP-graph based model to solve Resource-Constrained Project Scheduling Problems (RCPSP). The main drawback of this ACO-based model is related to the high number of pheromones created in the system. To overcome this issue we propose two adaptive Oblivion Rate heuristics to control the number of pheromones: the first one, called Dynamic Oblivion Rate, takes into account the overall number of pheromones produced in the system, whereas the second one inspires from the recently contributed Coral Reef Optimization (CRO) solver. A thorough experimental analysis has been carried out using the public PSPLIB library, and the obtained results have been compared to those of the most relevant contributions from the related literature. The performed experiments reveal that the Oblivion Rate heuristic removes at least 79% of the pheromones in the system, whereas the ACO algorithm renders statistically better results than other algorithmic counterparts from the literature.     

基于改进微粒群优化的学习路径优化控制方法

对学生学习的路径控制在智能化教学系统中是一个重要的问题。该文以知识空间理论为基础建立了学习状态空间,通过改进的微粒群算法对该学习状态空间的学习路径进行最优化控制,并利用死亡惩罚函数法把约束最优化学习路径问题转化成了无约束的最优化学习路径控制问题,引入交换子和交换序的概念对微粒群算法进行改进。在结果分析中,通过动态参数法,即动态变化交换子保留概率的方法提高微粒群的收敛效果,达到了最优化学习路径控制的目的。     

Using a Local Discovery Ant Algorithm for Bayesian Network Structure Learning

Bayesian networks (BNs) are knowledge representation tools capable of representing dependence or independence relationships among random variables. Learning the structure of BNs from datasets has received increasing attention in the last two decades, due to the BNs' capacity of providing good inference models and discovering the structure of complex domains. One approach for BNs' structure learning from data is to define a scoring metric that evaluates the quality of the candidate networks, given a dataset, and then apply an optimization procedure to explore the set of candidate networks. Among the most frequently used optimization methods for BN score-based learning is greedy hill climbing (GHC) search. This paper proposes a new local discovery ant colony optimization (ACO) algorithm and a hybrid algorithm max-min ant colony optimization (MMACO), based on the local discovery algorithm max-min parents and children (MMPC) and ACO to learn the structure of a BN. In MMACO, MMPC is used to construct the skeleton of the BN and ACO is used to orientate the skeleton edges, thus returning the final structure. The algorithms are applied to several sets of benchmark networks and are shown to outperform the GHC and simulated annealing algorithms.      

An analysis of students�� gaming behaviors in an intelligent tutoring system: predictors and impacts

Students who exploit properties of an instructional system to make progress while avoiding learning are said to be ??gaming?? the system. In order to investigate what causes gaming and how it impacts students, we analyzed log data from two Intelligent Tutoring Systems (ITS). The primary analyses focused on six college physics classes using the Andes ITS for homework and test preparation, starting with the research question: What is a better predictor of gaming, problem or student? To address this question, we developed a computational gaming detector for automatically labeling the Andes data, and applied several data mining techniques, including machine learning of Bayesian network parameters. Contrary to some prior findings, the analyses indicated that student was a better predictor of gaming than problem. This result was surprising, so we tested and confirmed it with log data from a second ITS (the Algebra Cognitive Tutor) and population (high school students). Given that student was more predictive of gaming than problem, subsequent analyses focused on how students gamed and in turn benefited (or not) from instructional features of the environment, as well as how gaming in general influenced problem solving and learning outcomes.     

一种学习稀疏BN最优结构的改进K均值分块学习算法

目前贝叶斯网络(Bayesian networks, BN)的传统结构学习算法在处理高维数据时呈现出计算负担过大、在合理时间内难以得到期望精度结果的问题.为了在高维数据下学习稀疏BN的最优结构, 本文提出了一种学习稀疏BN最优结构的改进K均值分块学习算法.该算法采用分而治之的策略, 首先采用互信息作为节点间距离度量, 利用融合互信息的改进K均值算法对网络分块; 其次, 使用MMPC (Max-min parent and children)算法得到整个网络的架构, 根据架构找到块间所有边的可能连接方向, 从而找到所有可能的图结构; 之后, 对所有图结构依次进行结构学习; 最终利用评分找到最优BN.实验证明, 相比现有分块结构学习算法, 本文提出的算法不仅习得了网络的精确结构, 且学习速度有一定提高; 相比非分块经典结构学习算法, 本文提出的算法在保证精度基础上, 学习速度大幅提高, 解决了非分块经典结构学习算法无法在合理时间内处理高维数据的难题.     

基于最大主子图分解的贝叶斯网络等价类学习算法

针对基于约束方法学习贝叶斯网络(BN)结构的不足,以及随着条件集的增大,利用统计方法进行条件独立(CI)测试不稳定等问题,提出一种基于最大主子图分解(MPD)的BN等价类学习算法.该算法首先通过MPD分解技术对BN的道德图进行分解;然后利用0阶和1阶CI测试识别部分子图中的V结构,对于初步未定的V结构利用局部评分搜索确定,从而避免了冗余检验,有效地减小了条件集的维数,并且提高了算法的效率.理论证明和实验结果均表明了所提出算法的有效性和合理性.     

类基于蚁群优化的贝叶斯置信网结构学习策略及性能分析*

针对贝叶斯置信网的结构学习问题,提出一种遵循典型ACO算法框架(ACO-TSP)的贝叶斯网结构学习算法（ACO-BN）,并拓展为包括EAS-BN、ACS-BN和MMAS-BN在内的一类算法。用这类算法在若干典型贝叶斯网络结构学习问题上分别与经典贝叶斯网学习算法(K2、B)、用于贝叶斯网学习的通用优化算法（simulated annealing、Tabu searching和genetic searching）以及L. M. de Campos等人提出的基于蚁群优化的贝叶斯网络结构学习算法 Ant-K2SN     

负载平衡路由：一种新的ACO路由策略*

通过改进ACO算法达到一种能实现通信网络负载平衡的群体智能路由策略。采用的主要方法为：将蚁群划分为若干个子群,不同子群的蚂蚁释放不同类型的信息素。通过不同类型信息素之间的相互制约作用,以及链路负载的测量,提出了三种策略实现负载平衡路由。在模拟网络上进行了不同策略的对比实验,以及与已有的群体智能路由算法的运行测试比较。实验结果表明,本文的路由策略具有较好的效果和一定的优势。     

BNC-VLA: bayesian network structure learning using a team of variable-action set learning automata

Bayesian Network (BN) is a probabilistic graphical model which describes the joint probability distribution over a set of random variables. One of the most important challenges in the field of BNs is to find an optimal network structure based on an available training dataset. Since the problem of searching the optimal BN structure belongs to the class of NP-hard problems, typically greedy algorithms are used to solve it. In this paper a learning automata-based algorithm has been proposed to solve the BNs structure learning problem. There is a learning automaton corresponding with each random variable and at each stage of the proposed algorithm, named BNC-VLA, a set of learning automata is randomly activated and determined the graph edges that must be appeared in that stage. Finally, the constructed network is evaluated using a scoring function. As BNC-VLA algorithm proceeds, the learning process focuses on the BN structure with higher scores. The convergence of this algorithm is theoretically proved; and also some experiments are designed to evaluate the performance of it. Experimental results show that BNC-VLA is capable of finding the optimal structure of BN in an acceptable execution time; and comparing against other search-based methods, it outperforms them.