Application of the probability-based covering algorithm model in text classification

The probability-based covering algorithm(PBCA) is a new algorithm based on probability distribution. It decides, by voting, the class of the tested samples on the border of the coverage area, based on the probability of training samples. When using the original covering algorithm(CA), many tested samples that are located on the border of the coverage cannot be classified by the spherical neighborhood gained. The network structure of PBCA is a mixed structure composed of both a feed-forward network and a feedback network. By using this method of adding some heterogeneous samples and enlarging the coverage radius,it is possible to decrease the number of rejected samples and improve the rate of recognition accuracy. Relevant computer experiments indicate that the algorithm improves the study precision and achieves reasonably good results in text classification.

Multiple-Instance Learning with Instance Selection via Constructive Covering Algorithm

Multiple-Instance Learning （MIL） is used to predict the unlabeled bags＇ label by learning the labeled positive training bags and negative training bags.Each bag is made up of several unlabeled instances.A bag is labeled positive if at least one of its instances is positive,otherwise negative.Existing multiple-instance learning methods with instance selection ignore the representative degree of the selected instances.For example,if an instance has many similar instances with the same label around it,the instance should be more representative than others.Based on this idea,in this paper,a multiple-instance learning with instance selection via constructive covering algorithm （MilCa） is proposed.In MilCa,we firstly use maximal Hausdorff to select some initial positive instances from positive bags,then use a Constructive Covering Algorithm （CCA） to restructure the structure of the original instances of negative bags.Then an inverse testing process is employed to exclude the false positive instances from positive bags and to select the high representative degree instances ordered by the number of covered instances from training bags.Finally,a similarity measure function is used to convert the training bag into a single sample and CCA is again used to classification for the converted samples.Experimental results on synthetic data and standard benchmark datasets demonstrate that MilCa can decrease the number of the selected instances and it is competitive with the state-of-the-art MIL algorithms.

FCV_1: A New Fast GreedyCovering Algorithm

In this paper, a new covering algorithm called FCV1 is presented. FCV1 comprises two algorithms, one of which is able to fast search for a partial rule and exclude the larg portion of negative examples, the other algorithm incorporates the more optimized greedy set-covering algorithm, and runs on a small portion of training examples. Hence,the training process of FCV1 is much faster than that of AQ15.

Multifractal analysis of complex networks

Complex networks have recently attracted much attention in diverse areas of science and technology. Many networks such as the WWW and biological networks are known to display spatial heterogeneity which can be characterized by their fractal dimensions. Multifractal analysis is a useful way to systematically describe the spatial heterogeneity of both theoretical and experimental fractal patterns. In this paper, we introduce a new box-covering algorithm for multifractal analysis of complex networks. This algorithm is used to calculate the generalized fractal dimensions Dq of some theoretical networks, namely scale-free networks, small world networks, and random networks, and one kind of real network, namely protein protein interaction networks of different species. Our numerical results indicate the existence of multifractality in scale-free networks and protein protein interaction networks, while the multifractal behavior is not clear-cut for small world networks and random networks. The possible variation of Dq due to changes in the parameters of the theoretical network models is also discussed.

MICkNN:Multi-Instance Covering kNN Algorithm

Mining from ambiguous data is very important in data mining. This paper discusses one of the tasks for mining from ambiguous data known as multi-instance problem. In multi-instance problem, each pattern is a labeled bag that consists of a number of unlabeled instances. A bag is negative if all instances in it are negative. A bag is positive if it has at least one positive instance. Because the instances in the positive bag are not labeled, each positive bag is an ambiguous. The mining aim is to classify unseen bags. The main idea of existing multi-instance algorithms is to find true positive instances in positive bags and convert the multi-instance problem to the supervised problem, and get the labels of test bags according to predict the labels of unknown instances. In this paper, we aim at mining the multi-instance data from another point of view, i.e., excluding the false positive instances in positive bags and predicting the label of an entire unknown bag. We propose an algorithm called Multi-Instance Covering kNN （MICkNN） for mining from multi-instance data. Briefly, constructive covering algorithm is utilized to restructure the structure of the original multi-instance data at first. Then, the kNN algorithm is applied to discriminate the false positive instances. In the test stage, we label the tested bag directly according to the similarity between the unseen bag and sphere neighbors obtained from last two steps. Experimental results demonstrate the proposed algorithm is competitive with most of the state-of-the-art multi-instance methods both in classification accuracy and running time.

RULES-IT： incremental transfer learning with RULES family

In today＇s world of excessive development in technologies, sustainability and adaptability of computer applications is a challenge, and future prediction became significant. Therefore, strong artificial intelligence （AI） became important and, thus, statistical machine learning （ML） methods were applied to serve it. These methods are very difficult to understand, and they predict the future without showing how. However, understanding of how machines make their decision is also important, especially in information system domain. Consequently, incremental covering algorithms （CA） can be used to produce simple rules to make difficult decisions. Nevertheless, even though using simple CA as the base of strong AI agent would be a novel idea but doing so with the methods available in CA is not possible. It was found that having to accurately update the discovered rules based on new information in CA is a challenge and needs extra attention. In specific, incomplete data with missing classes is inappropriately considered, whereby the speed and data size was also a concern, and future none existing classes were neglected. Consequently, this paper will introduce a novel algorithm called RULES-IT, in order to solve the problems of incremental CA and introduce it into strong AI. This algorithm is the first incremental algorithm in its family, and CA as a whole, that transfer rules of different domains to improve the performance, generalize the induction, take advantage of past experience in different domain, and make the learner more intelligent. It is also the first to introduce intelligent aspectsinto incremental CA, including consciousness, subjective emotions, awareness, and adjustment. Furthermore, all decisions made can be understood due to the simple representation of repository as rules. Finally, RULES-IT performance will be benchmarked with six different methods and compared with its predecessors to see the effect of transferring rules in the learning process, and to prove how RULES-IT actually solved the shortcoming of current incremental CA in addition to its improvement in the total performance.