Abstract: The Stochastic Point Location (SPL) problem Oommen is a fundamental learning problem that has recently found a lot of research attention. SPL can be summarized as searching for an unknown point in an interval under faulty feedback. The search is performed via a Learning Mechanism (LM) (algorithm) that interacts with a stochastic Environment which in turn informs it about the direction of the search. Since the Environment is stochastic, the guidance for directions could be faulty. The first solution to the SPL problem, which was pioneered two decades ago by Oommen, relies on discretizing the search interval and performing a controlled random walk on it. The state of the random walk at each step is considered to be the estimation of the point location. The convergence of the latter simplistic estimation strategy is proved for an infinite resolution, i.e., infinite memory. However, this strategy yields rather poor accuracy for low discretization resolutions. In this paper, we present two major contributions to the SPL problem. First, we demonstrate that the estimation of the point location can significantly be improved by resorting to the concept of mutual probability flux between neighboring states along the line. Second, we are able to accurately track the position of the optimal point and simultaneously show a method by which we can estimate the error probability characterizing the Environment. Interestingly, learning this error probability of the Environment takes place in tandem with the unknown location estimation. We present and analyze several experiments discussing the weaknesses and strengths of the different methods. PubDate: 2019-07-01

Abstract: Review texts, which have been shown helpful for recommending items for users, are often available in the form of user feedback for items. Despite the success of previous approaches exploring reviews for recommendations, they are all based on long review texts. The users reviews are, however, often short in real-world applications. In this paper, we develop a novel approach to leverage information from short review texts for recommendation based on word vector representations. We first build word vectors to represent items and users, which are called item-vector and user-vector, respectively. After that we concatenate item-vectors and user-vectors to form a set of training data with the rating scores that users give to items. Finally we train a regression model to predict the unknown rating scores. In our experiment, we show that our approach is effective, compared to state-of-the-art algorithms. PubDate: 2019-07-01

Abstract: For obtaining the more robust, novel, stable, and consistent clustering result, clustering ensemble has been emerged. There are two approaches in clustering ensemble frameworks: (a) the approaches that focus on creation or preparation of a suitable ensemble, called as ensemble creation approaches, and (b) the approaches that try to find a suitable final clustering (called also as consensus clustering) out of a given ensemble, called as ensemble aggregation approaches. The first approaches try to solve ensemble creation problem. The second approaches try to solve aggregation problem. This paper tries to propose an ensemble aggregator, or a consensus function, called as Robust Clustering Ensemble based on Sampling and Cluster Clustering (RCESCC).RCESCC algorithm first generates an ensemble of fuzzy clusterings generated by the fuzzy c-means algorithm on subsampled data. Then, it obtains a cluster-cluster similarity matrix out of the fuzzy clusters. After that, it partitions the fuzzy clusters by applying a hierarchical clustering algorithm on the cluster-cluster similarity matrix. In the next phase, the RCESCC algorithm assigns the data points to merged clusters. The experimental results comparing with the state of the art clustering algorithms indicate the effectiveness of the RCESCC algorithm in terms of performance, speed and robustness. PubDate: 2019-07-01

Abstract: In this paper, we focus on the task of estimating crowd count and high-quality crowd density maps. Among crowd counting methods, crowd density map estimation is especially promising because it preserves spatial information which makes it useful for both counting and localization (detection and tracking). Convolutional neural networks have enabled significant progress in crowd density estimation recently, but there are still open questions regarding suitable architectures. We revisit CNNs design and point out key adaptations, enabling plain a signal column CNNs to obtain high resolution and high-quality density maps on all major dense crowd counting datasets. The regular deep supervision utilizes the general ground truth to guide intermediate predictions. Instead, we build hierarchical supervisory signals with additional multi-scale labels to consider the diversities in deep neural networks. We begin by obtaining multi-scale labels based on different Gaussian kernels. These multi-scale labels can be seen as diverse representations in the supervision and can achieve high performance for better quality crowd density map estimation. Extensive experiments demonstrate that our approach achieves the state-of-the-art performance on the ShanghaiTech, UCF_CC_50 and UCSD datasets. PubDate: 2019-07-01

Abstract: In this paper, a many-objective evolutionary algorithm, named as a hybrid evolutionary algorithm based on knee points and reference vector adaptation strategies (KnRVEA) is proposed. Knee point strategy is used to improve the convergence of solution vectors. In the proposed algorithm, a novel knee adaptation strategy is introduced to adjust the distribution of knee points. KnRVEA is compared with five well-known evolutionary algorithms over thirteen benchmark test functions. The results reveal that the proposed algorithm provides better results than the others in terms of Inverted Generational Distance and Hypervolume. The computational complexity of the proposed algorithm is also analyzed. The statistical testing is performed to show the statistical significance of proposed algorithm. The proposed algorithm is also applied on three real-life constrained many-objective optimization problems to demonstrate its efficiency. The experimental results show that the proposed algorithm is able to solve many-objective real-life problems. PubDate: 2019-07-01

Abstract: The present paper aims to propose a new neural network called “sparse semi-autoencoder” to overcome the vanishing information problem inherent to multi-layered neural networks. The vanishing information problem represents a natural tendency of multi-layered neural networks to lose information in input patterns as well as training errors, including also natural reduction in information due to constraints such as sparse regularization. To overcome this problem, two methods are proposed here, namely, input information enhancement by semi-autoencoders and the separation of error minimization and sparse regularization by soft pruning. First, we try to enhance information in input patterns to prevent the information from decreasing when going through multi-layers. The information enhancement is realized in a form of new architecture called “semi-autoencoders”, in which information in input patterns is forced to be given to all hidden layers to keep the original information in input patterns as much as possible. Second, information reduction by the sparse regularization is separated from a process of information acquisition as error minimization. The sparse regularization is usually applied in training autoencoders, and it has a natural tendency to decrease information by restricting the information capacity. This information reduction in terms of the penalties tends to eliminate even necessary and important information, because of the existence of many parameters to harmonize the penalties with error minimization. Thus, we introduce a new method of soft pruning, where information acquisition of error minimization and information reduction of sparse regularization are separately applied without a drastic change in connection weights, as is the case of the pruning methods. The two methods of information enhancement and soft pruning try jointly to keep the original information as much as possible and particularly to keep necessary and important information by enabling the making of a flexible compromise between information acquisition and reduction. The method was applied to the artificial data set, eye-tracking data set, and rebel forces participation data set. With the artificial data set, we demonstrated that the selectivity of connection weights increased by the soft pruning, giving sparse weights, and the final weights were naturally interpreted. Then, when it was applied to the real data set of eye tracking, it was confirmed that the present method outperformed the conventional methods, including the ensemble methods, in terms of generalization. In addition, for the eye-tracking data set, we could interpret the final results according to the conventional eye-tracking theory of choice process. Finally, the rebel data set showed the possibility of detailed interpretation of relations between inputs and outputs. However, it was also found that the method had the limitation that the selectivity by the soft pruning could not be increased. PubDate: 2019-07-01

Abstract: In the past two decades, recommender systems have been successfully applied in many e-commerce companies. One of the promising techniques to generate personalized recommendations is collaborative filtering. However, it suffers from sparsity problem. Alleviating this problem, cross-domain recommender systems came into existence in which transfer learning mechanism is applied to exploit the knowledge from other related domains. While applying transfer learning, some information should overlap between source and target domains. Several attempts have been made to enhance the performance of collaborative filtering with the help of other related domains in cross-domain recommender systems framework. Although exploiting the knowledge from other domains is still challenging and open problem in recommender systems. In this paper, we propose a method namely User Profile as a Bridge in Cross-domain Recommender Systems (UP-CDRSs) for transferring knowledge between domains through user profile. Firstly, we build a user profile using demographical information of a user, explicit ratings and content information of user-rated items. Thereafter, the probabilistic graphical model is employed to learn latent factors of users and items in both domains by maximizing posterior probability. At last prediction on unrated item is estimated by an inner product of corresponding latent factors of users and items. Validating of our proposed UP-CDRSs method, we conduct series of experiments on various sparsity levels using cross-domain dataset. The results demonstrate that our proposed method substantially outperforms other without and with transfer learning methods in terms of accuracy. PubDate: 2019-07-01

Abstract: The generalized vertex cover problem (GVC) is a new variant of classic vertex cover problem which considers both vertex and weight of the edge into the objective function. The GVC is a renowned NP-hard optimization problem that finds the vertex subset where the sum of vertices and edge weight are minimized. In the mathematical field of electrical, networking and telecommunication GVC is used to solve the vertex cover problem. Finding the minimum vertex cover using GVC has a great impact on graph theory. Some exact algorithms were proposed to solve this problem, but they failed to solve it for real-world instances. Some approximation and metaheuristic algorithms also were proposed to solve this problem. Chemical Reaction Optimization (CRO) is an established population-based metaheuristic for optimization and comparing with other existing optimization algorithms it gives better results in most of the cases. The CRO algorithm helps to explore the search space locally and globally over the large population area. In this paper, we are proposing an algorithm by redesigning the basic four operators of CRO to solve GVC problem and an additional operator named repair function is used to generate optimal or near-optimal solutions. We named the proposed algorithm as GVC_CRO. Our proposed GVC_CRO algorithm is compared with the hybrid metaheuristic algorithm (MAGVCP), the local search with tabu strategy and perturbation mechanism (LSTP) and Genetic Algorithm (GA), which are state of the arts. The experimental results show that our proposed method gives better results than other existing algorithms to solve the GVC problem with less execution time in maximum cases. Statistical test has been performed to demonstrate the superiority of the proposed algorithm over the compared algorithm. PubDate: 2019-07-01

Abstract: Marginal Fisher analysis (MFA) is an efficient method for dimension reduction, which can extract useful discriminant features for image recognition. Since sparse learning can achieve better generalization ability and lessen the amount of computations in recognition tasks, this paper introduces sparsity into MFA and proposes a novel sparse modified MFA (SMMFA) method for facial expression recognition. The goal of SMMFA is to extract discriminative features by using the resulted sparse projection matrix. First, a modified MFA is proposed to find the original projection matrix. Similar to MFA, the modified MFA also defines the intra-class graph and the inter-class graph to describe geometry structure in the same class and local discriminant structure between different classes, respectively. In addition, the modified MFA removes the null space of the total scatter matrix. The sparse solution of SMMFA can be gained by solving the ℓ1 –minimization problem on the original projection matrix using the linearized Bregman iteration. Experimental results show that the proposed SMMFA can effectively extract intrinsic features and has better discriminant power than the state-of-the-art methods. PubDate: 2019-07-01

Abstract: Influence maximization, i.e. to maximize the influence spread in a social network by finding a group of influential nodes as small as possible, has been studied widely in recent years. Many methods have been developed based on either explicit Monte Carlo simulation or scoring systems, among which the former perform well yet are very time-consuming while the latter ones are efficient but sensitive to different spreading models. In this paper, we propose a novel influence maximization algorithm in social networks, named Reversed Node Ranking (RNR). It exploits the reversed rank information of a node and the effects of its neighbours upon this node to estimate its influence power, and then iteratively selects the top node as a seed node once the ranking reaches stable. Besides, we also present two optimization strategies to tackle the rich-club phenomenon. Experiments on both Independent Cascade (IC) model and Weighted Cascade (WC) model show that our proposed RNR method exhibits excellent performance and outperforms other state-of-the-arts. As a by-product, our work also reveals that the IC model is more sensitive to the rich-club phenomenon than the WC model. PubDate: 2019-07-01

Abstract: Malware is continuously evolving and becoming more sophisticated to avoid detection. Traditionally, the Windows operating system has been the most popular target for malware writers because of its dominance in the market of desktop operating systems. However, despite a large volume of new Windows malware samples that are collected daily, there is relatively little research focusing on Windows malware. The Windows Registry, or simply the registry, is very heavily used by programs in Windows, making it a good source for detecting malicious behavior. In this paper, we present RAMD, a novel approach that uses an ensemble classifier consisting of multiple one-class classifiers to detect known and especially unknown malware abusing registry keys and values for malicious intent. RAMD builds a model of registry behavior of benign programs and then uses this model to detect malware by looking for anomalous registry accesses. In detail, it constructs an initial ensemble classifier by training multiple one-class classifiers and then applies a novel swarm intelligence pruning algorithm, called memetic firefly-based ensemble classifier pruning (MFECP), on the ensemble classifier to reduce its size by selecting only a subset of one-class classifiers that are highly accurate and have diversity in their outputs. To combine the outputs of one-class classifiers in the pruned ensemble classifier, RAMD uses a specific aggregation operator, called Fibonacci-based superincreasing ordered weighted averaging (FSOWA). The results of our experiments performed on a dataset of benign and malware samples show that RAMD can achieve about 98.52% detection rate, 2.19% false alarm rate, and 98.43% accuracy. PubDate: 2019-07-01

Abstract: Human action recognition is an emerging goal of computer vision with several applications such as video surveillance and human-computer interaction. Despite many attempts to develop deep architectures to learn the spatio-temporal features of video, hand-crafted optical flow is still an important part of the recognition process. To engage the motion features deeply inside the learning process, we propose a spatio-temporal video recognition network where a motion-aware long short-term memory module is introduced to estimate the motion flow along with extracting spatio-temporal features. A specific optical flow estimator is subsumed which is based on kernelized cross correlation. The proposed network can be used without any extra learning process and there is no need to pre-compute and store the optical flow. Extensive experiments on two action recognition benchmarks verify the effectiveness of the proposed approach. PubDate: 2019-07-01

Abstract: Restricted Boltzmann machines (RBMs) can be trained by applying stochastic gradient ascent to the objective function as the maximum likelihood learning. However, it is a difficult task due to the intractability of marginalization function gradient. Several methodologies have been proposed by adopting Gibbs Markov chain to approximate this intractability including Contrastive Divergence, Persistent Contrastive Divergence, and Fast Contrastive Divergence. In this paper, we propose an optimization which is injecting noise to underlying Monte Carlo estimation. We introduce two novel learning algorithms. They are Noisy Persistent Contrastive Divergence (NPCD), and further Fast Noisy Persistent Contrastive Divergence (FNPCD). We prove that the NPCD and FNPCD algorithms benefit on the average to equilibrium state with satisfactory condition. We have performed empirical investigation of diverse CD-based approaches and found that our proposed methods frequently obtain higher classification performance than traditional approaches on several benchmark tasks in standard image classification tasks such as MNIST, basic, and rotation datasets. PubDate: 2019-07-01

Abstract: Fuzzy preference relation (FPR) is commonly used in solving multi-criteria decision making problems because of its efficiency in representing people’s perceptions. However, the FPR suffers from an intrinsic limitation of consistency in decision making. In this regard, many researchers proposed the consistent fuzzy preference relation (CFPR) as a decision-making approach. Nevertheless, most CFPR methods involve a traditional aggregation process which does not identify the interrelationship between the criteria of decision problems. In addition, the information provided by individual experts is indeed related to that provided by other experts. Therefore, the interrelationship of information on criteria should be dealt with. Based on this motivation, we propose a modified approach of CFPR with Geometric Bonferroni Mean (GBM) operator. The proposed method introduces the GBM as an operator to aggregate information. The proposed method is applied to a case study of assessing the quality of life among the population in Setiu Wetlands. It is shown that the best option derived by the proposed method is consistent with that obtained from the other methods, despite the difference in aggregation operators. PubDate: 2019-07-01

Abstract: Recent advances in automatic machine learning (aML) allow solving problems without any human intervention. However, sometimes a human-in-the-loop can be beneficial in solving computationally hard problems. In this paper we provide new experimental insights on how we can improve computational intelligence by complementing it with human intelligence in an interactive machine learning approach (iML). For this purpose, we used the Ant Colony Optimization (ACO) framework, because this fosters multi-agent approaches with human agents in the loop. We propose unification between the human intelligence and interaction skills and the computational power of an artificial system. The ACO framework is used on a case study solving the Traveling Salesman Problem, because of its many practical implications, e.g. in the medical domain. We used ACO due to the fact that it is one of the best algorithms used in many applied intelligence problems. For the evaluation we used gamification, i.e. we implemented a snake-like game called Traveling Snakesman with the MAX–MIN Ant System (MMAS) in the background. We extended the MMAS–Algorithm in a way, that the human can directly interact and influence the ants. This is done by “traveling” with the snake across the graph. Each time the human travels over an ant, the current pheromone value of the edge is multiplied by 5. This manipulation has an impact on the ant’s behavior (the probability that this edge is taken by the ant increases). The results show that the humans performing one tour through the graphs have a significant impact on the shortest path found by the MMAS. Consequently, our experiment demonstrates that in our case human intelligence can positively influence machine intelligence. To the best of our knowledge this is the first study of this kind. PubDate: 2019-07-01

Abstract: In this paper, a crack detection mechanism for concrete tunnel surfaces is presented. The proposed methodology leverages deep Convolutional Neural Networks and domain-specific heuristic post-processing techniques to address a variety of challenges, including high accuracy requirements, low operational times and limited hardware resources, poor and variable lighting conditions, low textured lining surfaces, scarcity of training data, and abundance of noise. The proposed framework leverages the representational power of the convolutional layers of CNNs, which inherently selects effective features, thus obviating the need for the tedious task of handcrafted feature extraction. Additionally, the good performance rates attained by the proposed framework are acquired at a significantly lower execution time compared to other techniques. The presented mechanism was designed and developed as a core component of an autonomous robotic inspector deployed and validated in the tunnels of Egnatia Motorway in Metsovo, Greece. The obtained results denote the proposed approach’s superiority over a variety of methods and suggest a promising potential as a driver of autonomous concrete-lining tunnel-inspection robots. PubDate: 2019-07-01

Abstract: Feature extraction is a crucial technique for data preprocessing in classification tasks such as protein classification and image classification. Datasets with tree class hierarchies have become extremely common in many practical classification tasks. However, existing flat feature extraction algorithms tend to assume that classes are independent and ignore the hierarchical information of class structure within a dataset. In this paper, we propose a hierarchical feature extraction algorithm based on discriminant analysis (HFEDA). HFEDA first decomposes the highly complex feature extraction problem into smaller problems by creating sub-datasets for non-leaf nodes according to the tree class hierarchy of dataset. Secondly, different from flat algorithms, HFEDA takes the hierarchical class structure into account in dimensionality reduction process, and calculates the projection matrices for the non-leaf nodes in the tree class hierarchy. In this way, HFEDA can just focus on discriminating the several categories under the same parent node. Finally, HFEDA does not need to determine the optimal feature subset size, which is challenging for most feature selection algorithms. Extensive experiments on different type datasets and typical classifiers demonstrate the effectiveness and efficiency of the proposed algorithm. PubDate: 2019-07-01

Abstract: Fuzzy linguistic approach is considered as an effective solution to accommodate uncertainties in qualitative decision making. In the face of increasingly complicated environment, we usually face the fact that multi-attribute group decision making contains nested information, and the whole process needs to be evaluated twice so that the experts can make full use of decision information to get more accurate result, and we call this kind of problem as two-stage multi-attribute group decision making (TSMAGDM). However, the existing linguistic approaches cannot represent such nested evaluation information to deal with the above situation. In this paper, a novel linguistic expression tool called nested probabilistic-numerical linguistic term set (NPNLTS) which considers both quantitative and qualitative information, is proposed to handle TSMAGDM. Based on which, some basic operational laws and aggregation operators are put forward. Then, an aggregation method and an extended TOPSIS method are developed respectively in TSMAGDM with NPNLTSs. Finally, an application case about strategy initiatives of HBIS GROUP on Supply-side Structural Reform is presented, and some analyses and comparisons are provided to validate the proposed methods. PubDate: 2019-07-01

Abstract: In this study, a hybrid and layered Intrusion Detection System (IDS) is proposed that uses a combination of different machine learning and feature selection techniques to provide high performance intrusion detection in different attack types. In the developed system, firstly data preprocessing is performed on the NSL-KDD dataset, then by using different feature selection algorithms, the size of the dataset is reduced. Two new approaches have been proposed for feature selection operation. The layered architecture is created by determining appropriate machine learning algorithms according to attack type. Performance tests such as accuracy, DR, TP Rate, FP Rate, F-Measure, MCC and time of the proposed system are performed on the NSL-KDD dataset. In order to demonstrate the performance of the proposed system, it is compared with the studies in the literature and performance evaluation is done. It has been shown that the proposed system has high accuracy and a low false positive rates in all attack types. PubDate: 2019-07-01

Abstract: The objective of the color quantization problem is to reduce the number of different colors of an image, in order to obtain a new image as similar as possible to the original. This is a complex problem and several solution techniques have been proposed to solve it. Among the most novel solution methods are those that apply swarm-based algorithms. These algorithms define an interesting solution approach, since they have been successfully applied to solve many different problems. This paper presents a color quantization method that combines the Artificial Bee Colony algorithm with the Ant-tree for Color Quantization algorithm, creating an improved version of a previous method that combines artificial bees with the K-means algorithm. Computational results show that the new method significantly reduces computing time compared to the initial method, and generates good quality images. Moreover, this new method generates better images than other well-known color quantization methods such as Wu’s method, Neuquant, Octree or the Variance-based method. PubDate: 2019-07-01