The importance of learning capability for the realization of flexible information processing is being recognized lately, and active research in the area of machine learning is being pursued from various approaches. Among them is the field of computational learning theory, whose goal is to formally model learning activities and analyze the complexity of learning problems in the learning models thus set forth, both from computational and statistical points of view. This article will review some of the new learning models in this field as well as the results obtained concerning them. In particular, we focus on the learning models introduced and studied by researchers at NEC C&C Research Laboratories.

View full abstract

Software reliability is one of the most important characteristics of software quality. Its measurement and management technologies during the software life-cycle are essential to produce and maintain quality/reliable software systems. In this paper, we discuss software reliability modeling to describe a software fault-detection or software failure-occurrence phenomenon during the testing and operation phases and to assess software reliability quantitatively. A software failure is defined as an unacceptable departure of the program operation caused by software faults introduced in the development process. In particular, hazard rate models and NHPP(non-homogeneous Poisson process) models are investigated as useful software reliability growth models which formulate the relationship between the cumulative number of faults detected by the testing(or the time-interval between software failures) and the time span of the testing. Two software management problems are also discussed as application of software reliability models : Statistical testing-progress control and maintenance cost optimization. Further, numerical illustrations for the software reliability analysis and the software management problems are presented.

View full abstract