Changqin Huang

An examination of on-line machine learning approaches for pseudo-random generated data

A pseudo-random generator is an algorithm to generate a sequence of objects determined by a truly random seed which is not truly random. It has been widely used in many applications, such as cryptography and simulations. In this article, we examine current popular machine learning algorithms with various on-line algorithms for pseudo-random generated data in order to find out which machine learning approach is more suitable for this kind of data for prediction based on on-line algorithms. To further improve the prediction performance, we propose a novel sample weighted algorithm that takes generalization errors in each iteration into account. We perform intensive evaluation on real Baccarat data generated by Casino machines and random number generated by a popular Java program, which are two typical examples of pseudo-random generated data. The experimental results show that support vector machine and k-nearest neighbors have better performance than others with and without sample weighted algorithm in the evaluation data set.

NGD: Filtering Graphs for Visual Analysis

Graph visualization finds wide applications in different areas. As the popularity of social network sites is increasing, it becomes particularly useful in visual analysis of these sites. A number of algorithms for graph visualization have been developed over the past decades. The issue on how to design and develop algorithms by taking into account the characteristics of real graphs such as scale-free and hierarchical structures, however, has not been well addressed. In this paper, we extend the concept of a node degree to a node global degree for a node in a graph, and present an algorithm that computes their scores of all nodes. By taking advantage of the common structure features of real networks, two scalable extensions of this algorithm are further provided that are able to approximate computation results. Based on node global degrees, a filtering approach is presented to reduce the visual complexity of a layout. Extensive experiments have demonstrated the performance of the proposed algorithms in terms of two common evaluation metrics, as well as visualization results. In addition, we have implemented the algorithms in a prototype system, which enable users to explore a graph at continuous levels of details in real time, as evidenced by several real examples.

An Improved MOEA/D Based on Reference Distance for Software Project Portfolio Optimization

As it is becoming extremely competitive in software industry, large software companies have to select their project portfolio to gain maximum return with limited resources under many constraints. Project portfolio optimization using multiobjective evolutionary algorithms is promising because they can provide solutions on the Pareto-optimal front that are difficult to be obtained by manual approaches. In this paper, we propose an improved MOEA/D (multiobjective evolutionary algorithm based on decomposition) based on reference distance (MOEA/D_RD) to solve the software project portfolio optimization problems with optimizing 2, 3, and 4 objectives. MOEA/D_RD replaces solutions based on reference distance during evolution process. Experimental comparison and analysis are performed among MOEA/D_RD and several state-of-the-art multiobjective evolutionary algorithms, that is, MOEA/D, nondominated sorting genetic algorithm II (NSGA2), and nondominated sorting genetic algorithm III (NSGA3). The results show that MOEA/D_RD and NSGA2 can solve the software project portfolio optimization problem more effectively. For 4-objective optimization problem, MOEA/D_RD is the most efficient algorithm compared with MOEA/D, NSGA2, and NSGA3 in terms of coverage, distribution, and stability of solutions.

A novel multiple layers name disambiguation framework for digital libraries using dynamic clustering

In many types of databases, such as a science bibliography database, the name attribute is the most commonly used identifier to recognize entities. However, names are frequently ambiguous and not always unique, thereby causing problems in various fields. Name disambiguation is a data management task that aims to properly distinguish different entities that share the same name, particularly for large databases such as digital libraries, because the information that can be used to identify author’s name is limited. In digital libraries, the issue of ambiguous author names occurs due to the existence of multiple authors with the same name or different name variations for the same author. Most previous works conducted to solve this issue frequently used hierarchical clustering approaches based on information within citation records, e.g., co-authors and publication titles. In the present study, we propose a multiple layers name disambiguation framework that is not only applicable to digital libraries but can also be easily extended to other applications. Our framework adopts a dynamic clustering mechanism to minimize clustering errors. We evaluated our approach on real world corpora, and favorable experiment results indicated that our proposed framework was feasible.

Large-scale semantic web image retrieval using bimodal deep learning techniques

Abstract Semantic web image retrieval is useful to end-users for semantic image searches over the Internet. This paper aims to develop image retrieval techniques for large-scale web image databases. An advanced retrieval system, termed Multi-concept Retrieval using Bimodal Deep Learning (MRBDL), is proposed and implemented using Convolutional Neural Networks (CNNs) which can effectively capture semantic correlations between a visual image and its free contextual tags. Different from existing approaches using multiple and independent concepts in a query, MRBDL considers multiple concepts as a holistic scene for retrieval model learning. In particular, we first use a bimodal CNN to train a holistic scene classifier in two modalities, and then semantic correlations of the sub-concepts included in the images are leveraged to boost holistic scene recognition. The predicted semantic scores obtained from holistic scene classifier are combined with complementary information on web images to improve the retrieval performance. Experiments have been carried out over two publicly available web image databases. The results show that our proposed approach performs favorably compared with several other state-of-the-art methods.

A novel approach for entity resolution in scientific documents using context graphs

Abstract Entity resolution refers to disambiguating and resolving entities in structured and unstructured data. Developments of effective resolution algorithms are significant for processing scientific documents, particularly for biomedical literature. Specifically, name ambiguity among biomedical entities is a primary task that needs to be solved in the knowledge extraction process. In this paper, we present a novel approach to disambiguating gene/protein names by using context graphs. A set of abstracts of documents is used to build the context graphs through disclosing the indirect co-occurrence relationships among words. Feature vectors of the graphs can be constructed according to information gain (IG) on the word set. To evaluate the IG values, we propose a new metrics that integrates the word frequency (WF), dispersion degree (DD) and concentration degree (CD). Finally, entity resolution is performed by applying a support vector machine (SVM). Compared to existing approaches, the proposed method is capable of discovering latent information from the context of entity names, rather than using some statistical information such as the number of occurrences of words. Based on the results from comprehensive experiments over two benchmark datasets, we conclude that our proposed method, compared to several existing solutions, for resolving ambiguity entities is promising.

Improved expert selection model for forex trading

Online prediction is a process that repeatedly predicts the next element in the coming period from a sequence of given previous elements. This process has a broad range of applications in various areas, such as medical, streaming media, and finance. The greatest challenge for online prediction is that the sequence data may not have explicit features because the data is frequently updated, which means good predictions are difficult to maintain. One of the popular solutions is to make the prediction with expert advice, and the challenge is to pick the right experts with minimum cumulative loss. In this research, we use the forex trading prediction, which is a good example for online prediction, as a case study. We also propose an improved expert selection model to select a good set of forex experts by learning previously observed sequences. Our model considers not only the average mistakes made by experts, but also the average profit earned by experts, to achieve a better performance, particularly in terms of financial profit. We demonstrate the merits of our model on two real major currency pairs corpora with extensive experiments.

Online Prediction for Forex with an Optimized Experts Selection Model

Online prediction is a process to repeatedly predict the next element from a sequence of given previous elements. It has a broad range of applications on various areas, such as medical and finance. The biggest challenge of online prediction is sequence data does not have explicit features, which means it is difficult to remain good predictions. One of popular solution is to make prediction with expert advice, and the challenge is to pick the right experts with minimum cumulative loss. In this article, we use forex prediction as a case study, and propose a model that can select a good set of forex experts by learning a set of previous observed sequences. To achieve better performance, our model not only considers the average mistakes made by experts but also takes the average profit earn by experts into account. We demonstrate the merits of our model on a real major currency pairs data set.

Context-based prediction for road traffic state using trajectory pattern mining and recurrent convolutional neural networks

Abstract With the broad adoption of the global positioning system-enabled devices, the massive amount of trajectory data is generated every day, which results in meaningful traffic patterns. This study aims to predict the future state of road traffic instead of merely showing the current traffic condition. First, we use a clustering method to group similar trajectories of a particular period together into a cluster for each road. Second, for each cluster, we average the lengths and angles of the entire trajectories in the group as the representative trajectory, which is regarded as the road pattern. Third, we create a feature vector for each road based on its historical traffic conditions and neighbor road patterns. Finally, we design a recurrent convolutional neural network for modeling the complex nonlinear relationship among features to predict road traffic conditions. Experimental results show that our approach performs more favorably compared with several traditional machine learning and state-of-the-art algorithms.

Combining Convolutional Neural Network and Markov Random Field for Semantic Image Retrieval

With the rapidly growing number of images over the Internet, efficient scalable semantic image retrieval becomes increasingly important. This paper presents a novel approach for semantic image retrieval by combining Convolutional Neural Network (CNN) and Markov Random Field (MRF). As a key step, image concept detection, that is, automatically recognizing multiple semantic concepts in an unlabeled image, plays an important role in semantic image retrieval. Unlike previous work that uses single-concept classifiers one by one, we detect semantic multiconcept by using a multiconcept scene classifier. In other words, our approach takes multiple concepts as a holistic scene for multiconcept scene learning. Specifically, we first train a CNN as a concept classifier, which further includes two types of classifiers: a single-concept fully connected classifier that is best suited to single-concept detection and a multiconcept scene fully connected classifier that is good for holistic scene detection. Then we propose an MRF-based late fusion approach that is able to effectively learn the semantic correlation between the single-concept classifier and multiconcept scene classifier. Finally, the semantic correlation among the subconcepts of images is cought to further improve detection precision. In order to investigate the feasibility and effectiveness of our proposed approach, we conduct comprehensive experiments on two publicly available image databases. The results show that our proposed approach outperforms several state-of-the-art approaches.