Abstract Convolutional neural networks (CNNs), a representative type of deep neural networks, are used in various fields. There are problems that should be solved to operate CNN in the real-world. In real-world operating environments, the CNN’s performance may be degraded due to data of untrained types, which limits its operability. In this study, we propose a method for identifying data of a type that the model has not trained on based on the neuron cluster, a set of neurons activated based on the type of input data. In experiments performed on the ResNet model with the MNIST, CIFAR-10, and STL-10 datasets, the proposed method identifies data of untrained and trained types with an accuracy of 85% or higher. The more data used for neuron cluster identification, the higher the accuracy; conversely, the more complex the dataset's characteristics, the lower the accuracy. The proposed method uses only the information of activated neurons without any addition or modification of the model’s structure; hence, the computational cost is low without affecting the classification performance of the model.

State diagrams commonly used for describing dynamic behavior of systems or objects can have various forms that specify the same behavior. Highly understandable state diagrams are important for efficient and clear communication between developers. Measurement of understandability of state diagrams is the first step to obtain highly understandable ones. Unfortunately, there is little research on metrics for evaluating state diagram's understandability. In this paper, we propose cohesion and coupling metrics, named Average Cohesiveness Of States (ACOS) and Average number of Similar States Of States (ASSOS), to evaluate understandability of state diagrams. To validate effectiveness of the metrics, we compared understandability of three state diagrams of different representations but the same behavior for two systems. The experiments showed that ACOS and ASSOS can capture understandability more correctly than existing structure-based metrics.

Abstract Refactoring is a widely accepted technique to improve the software quality by restructuring its design without changing its behavior. In general, a sequence of refactorings needs to be applied until the quality of the code is improved satisfactorily. In this case, the final design after refactoring can vary with the application order of refactorings, thereby producing different quality improvements. Therefore, it is necessary to determine a proper refactoring schedule to obtain as many benefits as possible. However, there is little research on the problem of generating appropriate schedules to maximize quality improvement. In this paper, we propose an approach to automatically determine an appropriate schedule to maximize quality improvement through refactoring. We first detect code clones that are suitable for refactoring and generate the most beneficial refactoring schedule to remove them. It is straightforward to select the best from the exhaustively enumerated schedules. However, such a technique becomes NP‐hard, as the number of available refactorings increases. We apply a genetic algorithm (GA) to generate the best refactoring schedule within a reasonable time to cope with this problem. We compare the GA‐based approach with manual scheduling, greedy heuristic‐based, and exhaustive approaches for four open systems. The results show that the proposed GA‐based approach generates more beneficial schedules than the others. Copyright © 2010 John Wiley & Sons, Ltd.

Abstract In recent years, Radio Frequency Identification (RFID) industries have taken a great interest in utilizing the benefits of RFID for supply chain management, inventory control and various other applications. This paper proposed an adaptive load balancing technique for RFID middleware systems to meet the demands of scalability and heterogeneity. First, we explored five basic load balancing policies, namely, information policy, job selection policy, transfer policy, initiation policy and location policy. Eighteen load balancing schemes were then proposed for RFID middleware systems that were combinations of various types of the five basic load balancing policies. Our empirical study suggested that these load balancing strategies performed differently under different workload statuses. Finally, an adaptive load balancing strategy was proposed. The load balancing schemes and the proposed adaptive load balancing strategy have been implemented in the RFID Middleware Load Management System (RM‐LMS). Copyright © 2010 John Wiley & Sons, Ltd.

This paper proposes an approach to checking behavioral compatibility between Web services. If Web service B can be used in replacement of Web service A in such a way that the replacement is transparent to clients, Web service B is compatible to Web service A. We use state machines with guarded transitions to specify behaviors of Web services. To check compatibility between two Web services, we propose an extended version of the conventional methods rule, which has been used in object-oriented paradigm. To support our approach, we have implemented a tool. First, the tool constructs a state machine for a Web service whose behavior is expressed in WSDL and WSCI. Then, the tool can verify compatibility between Web services by using the extended methods rule.

The methodologies of product line engineering emphasize proactive reuse to construct high-quality products more quickly that are less costly. Requirements engineering for software product families differs significantly from requirements engineering for single software products. The requirements for a product line are written for the group of systems as a whole, with requirements for individual systems specified by a delta or an increment to the generic set. Therefore, it is necessary to identify and explicitly denote the regions of commonality and points of variation at the requirements level. In this paper, we suggest a method of producing requirements that will be a core asset in the product line. We describe a process for developing domain requirements where commonality and variability in a domain are explicitly considered. A CASE environment, named DREAM, for managing commonality and variability analysis of domain requirements is also described. We also describe a case study for an e-travel system domain where we found that our approach to developing domain requirements based on commonality and variability analysis helped to produce domain requirements as a core asset for product lines.

The existing cohesion metrics for classes do not consider the characteristics of dependent instance variables that are commonly used in a class and, thus, do not properly reflect the cohesiveness of the class. This paper presents an approach for improving the cohesion metrics by considering the characteristics of the dependent instance variables in an object-oriented program.

Abstract The authors insist that monotonicity is a necessary property of a good cohesion metric and the violation of the monotonicity property limits the application of CBMC. They also state that the augmented CBMC can also be used as a guideline for quality evaluation and restructuring of poorly designed classes. This paper raises the question about the necessity of monotonicity by analyzing the reason that causes CBMC to violate the monotonicity property. In addition, we give a detailed description of the restructuring procedure based on CBMC. Copyright © 2003 John Wiley & Sons, Ltd.

Abstract Indicating that CBMC does not satisfy the monotonic property in terms of the number of interactions, Xu and Zhou proposed an augmented definition of CBMC by adopting cut set instead of glue methods. The augmented CBMC clearly satisfies the monotonic property. However, CBMC is designed to overcome the problem with respect to the number of interactions and, therefore, focuses on the interaction pattern, especially, member connectivity. Consequently, it does not make sense to mention the monotonic property of CBMC with respect to interaction number. Moreover, the notion of glue methods allows several interpretations on the design quality of a class. However, that meaningful interpretation is not possible for the augmented definition due to the removal of the notion of glue methods. Copyright © 2001 John Wiley & Sons, Ltd.

In object-oriented systems, cohesion refers to the degree of the relatedness of the members in a class and strong cohesion has been recognized as a highly desirable property of classes. We note that the existing cohesion measures do not take into account some characteristics of classes, and thus often fail to properly reflect the cohesiveness of classes. To cope with such a problem, we propose a new cohesion measure where the characteristics of classes are incorporated. Our cohesion measure takes into account the members that actually have impact on the cohesiveness of a class, and is defined in terms of the degree of the connectivity among those members. We develop a cohesion measurement tool for C++ programs, and perform a case study on a well-known class library in order to demonstrate the effectiveness of our new measure. By performing principal component analysis, we also demonstrate that our measure captures a new aspect of class properties which is not captured by the existing cohesion measures. Copyright © 2000 John Wiley & Sons, Ltd.

Modified programs are required to be retested to determine whether changes have been made correctly and to investigate whether those cause any adverse effect on the program behavior. Then, the key issue is the extent of the retesting tasks. To save time and efforts, retesting should be done on only the parts that might be affected by changes. Change impact analysis is the activity which identifies the parts to be retested. Object-oriented paradigm is based on several concepts such as encapsulation, inheritance, polymorphism, and dynamic binding. Although these features contribute to the reusability and extensibility of systems, they produce complex dependencies between classes and attributes, which make it difficult to identify the affected components by changes. We propose an approach for analyzing change impact in a class hierarchy. Our approach is based on the class firewall method, but aims to reduce the retesting effort significantly by taking a member function as a test unit. We deal with the impacts of changes associated with various object-oriented features as well as types of changes which can occur at the level of a data member a member function, a class, and an inheritance relation.

Classes are the fundamental concepts in the object-oriented paradigm. They are the basic units of object-oriented programs, and serve as the units of encapsulation, which promotes their modifiability and reusability. However, improper modeling in the design phase or uncontrolled changes during the maintenance phase can degrade the quality of classes, which leads to programs which are cumbersome to maintain or extend. The paper proposes a method for assessing the quality of classes and restructuring them. The authors define an abstraction as the most cohesive form of a class, and the cohesion indicates the extent to which a class approaches the abstraction. Assessment and restructuring are both based on the notion of class cohesion. Less cohesive classes can be transformed into a form of an abstraction by restructuring them.