Recent years, with the widespread usage of large code pre-trained model, the application of deep learning in Software Engineering (SE) tasks has made significant progress. To achieve better performance, it is important to capture the code features during pre-training process. The existing studies have made many attempts for this challenge, such as trying to introduce Syntax Tree (AST) or data flow graph into pre-training process. However, existing studies still have limitations in the comprehensiveness and simplicity of capturing code features. Knowledge graphs (KGs), which represent entities and their relationships in graph structures, have been extensively applied in intelligent systems for domain-specific tasks. Inspired by this, we attempted a knowledge graph guided code pre-training approach in this study, hoping that the codespecific knowledge graph could help code pre-trained model achieve better code understanding ability. In our approach, we first proposed a new type of code-specific knowledge graph which integrates syntactic, structural, and semantic information from a code fragment. Next, we extract the code feature triplets and code concept triplets from the code knowledge graph as part of input data. We have also designed a specialized pre-training task to enhance the pre-trained model’s code feature comprehension ability. To evaluate our approach, we trained a baseline pre-trained model (KG-Code) and validated its effectiveness on two code understanding downstream tasks: code summarization and code clone detection. The experimental results show that introducing knowledge graph as guidance in the pre-training process could efficiently improve the code understanding ability.

The success of nnU-Net proves the significance of the rationality of workflow architecture and configuration settings in improving segmentation accuracy. However, since that, most efforts to improve U-Net have continued to address CNN inner limitations caused by architecture. These methods encountered challenges such as limited generalization, difficulty in managing data with varying distribution patterns. To tackle these issues, We designed a standardized processing workflow specifically tailored for the convolutional layers of U-Net : Arena Mechanism (ARM), inspired by game theory, which encompasses three stages: Recruit, Train and Fight. We augment the convolutional layers of U-Net with two additional processing branches (Recruit), serving as "challengers." These challengers are first optimized within a Hybrid Adaptive Weighting module to refine the feature representation of each internal channel (Train). During forward propagation, we design a cooperative loss and reward function to determine the optimal confidence distribution and fusion strategy for branch outputs (Fight). This standardized mechanism is designed to further nnU-Net’s core principle of "Automatic Adaptation." Experiments have demonstrated that this approach achieves state-of-the-art (SOTA) performance on both the ACDC, BraTS21 and KiTS datasets.

Multimodal Aspect-Category-Sentiment Triple Extraction (MACSTE) is a highly complex subtask within Multimodal Aspect-Based Sentiment Analysis (MABSA), requiring simultaneous attribute extraction and sentiment polarity prediction from image-text pairs. While existing research often emphasizes modality fusion and alignment, it frequently neglects the design of information flow pathways, leading to suboptimal utilization of complementary information. Additionally, modality-specific noise may compromise the robustness and accuracy of multimodal classification, with traditional filtering methods often degrading data quality. To overcome these challenges, we propose the Metabolic Attention Network Utilizing Uncertainty Estimation (MetaCert). MetaCert integrates two key components: the Metabolic Attention Mechanism (MAM), inspired by bio-chemical metabolic networks and enhanced by cross-attention for improved information exchange; and the Uncertainty Estimation Network (UEN), which optimizes the semantic contributions of each modality while preserving data integrity, thereby enhancing classification accuracy. Our approach achieves state-of-the-art (SOTA) results on the TWITTER-15 and TWITTER-17 datasets.

Family-based model checking of product families is a technique to verify multiple products simultaneously by model checking the variational model. This technique is useful for design verification of core assets, etc. but it has the problem that scalability issues are more pronounced compared to model checking of a single system. Although it would be more convenient to use existing model checkers, most studies use dedicated model checkers because of the efficiency. In this paper, we propose a method of family-based model checking using an existing probabilistic model checker, PRISM, for FTS (Featured Transition Systems). It is possible to verify whether there exist feature configurations for which the property holds, and also obtain a specific feature configuration example for which the property holds. In order to improve scalability, we divide the corresponding FD (Feature Diagram) into smaller FDs by adding constraints on the selection or deselection of features and perform model checking based on these divided FDs independently. Although this method does not always guarantee improvement in theory, our evaluation shows that it can be quite effective in many cases by selecting features to be used for division using metrics based on the probability of occurrence of the features.

In Software Product Line (SPL) development, one of promising techniques for core asset testing is to test a subset of SPL as representative products. SPL pairwise testing is a such technique in which each product corresponds to a possible feature configuration in the feature model (FM) and representative products are selected so as to all possible feature pairs are included. It is also important to prioritize representative products, because it could improve the effectiveness of core asset testing especially when the testing resource is limited. In this paper, we propose a prioritization method for SPL pairwise testing based on user profiles. A user profile is a set of user groups and their occurrence probabilities such as the percentages of user groups in a market that use specific devices, applications or services. These profiles are used as the probabilities of feature choices at decision points such as optional features and alternative features in a FM. Based on that, we calculate the probability for obtaining a feature pairs (PFP for short), and generate representative products with priority. Most researches relate to the probabilities about FM handle the probability for obtaining a single feature (PSF for short). Based on PSF, we could estimate PFP. However, this estimation is not appropriate for the prioritization especially when conditional probabilities appear in user profiles. In our method, we directly calculate PFP and determine the priorities. We evaluate the method to show advantages of prioritizations using PFP over those using PSF, and also analyze the characteristics of the method.

In recent years, because web applications have been handling increasingly important processing tasks, it is ever more important to avoid errors.Model checking is one verification method for detecting errors, whereby it is necessary to model the web application in order to verify it.However, typical web application developers may lack knowledge on creating the verification model.Furthermore, web applications have become increasingly diversified owing to web-browsing technological advancements and other factors.Among these, the single-page application (SPA) using a component-based web application framework, such as Angular, has attracted attention because of its excellent user experience.However, it makes modeling more difficult since the page structure is complicated by the intricate combination of components.In this paper, we therefore present a method to automatically generate verification models from source code and perform model checking.The method enables verification of SPA page transitions using the component-based framework.We apply our implemented automated tool to several applications.First, we experiment using sample applications that do not inject bugs and others that intentionally inject bugs.Moreover, we apply the method to real applications published on the Internet.The desired results are obtained, thereby confirming that the proposed method is effective.

Advances in device technology have promoted the development of small appliances such as wearable devices, IoT equipment, and small home electronics equipment. Considering a product family of such device/equipment, each product in the family may have different hardware (micro controller unit, devices such as sensors and actuators, and different configurations). Software for such equipment tends to be compact and handles the devices directly via drivers. However, software engineers are not familiar with driver development and this causes development bottleneck. Therefore, in order to develop products efficiently, derivations of software and hardware are not enough, i.e., the systematic derivation of drivers is indispensable. In this paper, we propose a system product line development method for small appliances in which drivers are systematically derived. In this method, hardware features and software features are managed in terms of feature models, and system products are derived by specifying features of hardware and software. Based on these, drivers are systematically derived. This paper proposes 1) an extension of the UML/MARTE model that can represent hardware information that is necessary for driver generation, and 2) a systematic driver derivation method based on variability management and model-driven engineering techniques. We evaluate the method using an example of a motion tracking system product family.

Many of the benefits expected from software product lines (SPL) [1-2] are based on the assumption that the additional investment required for domain engineering, pays off during application engineering when products are derived from the product line [3]. However, to fully exploit this we need to optimize application engineering processes and handle the reusable artifacts of an SPL in a systematic and efficient manner. In this context, the joint MAPLE/SCALE workshop focuses on two closely related aspects: how model-driven approaches can help to achieve systematic and efficient derivation of products and how scalability challenges can be addressed that arise from the application of SPL techniques to SPLs of realistic size and complexity. The workshop aims to explore and explicate the current status and ongoing work in model-driven approaches and/or scalability of SPLs and the transfer of knowledge between different disciplines and application domains.

In this paper, we first reveal an important problem called a state inconsistency problem among tasks based on a design model of a CD/DVD systems.This problem is not only of CD/DVD systems but also of typical embedded software because it may occur in a typical structure that such software has. Thus, we propose a method to detect the problem in the design model with a model checking tool and to derive behavior in which it does not occur.

Our society depend on embedded and ubiquitous computing and the reliability of embedded software becomes more and more important. We have conducted a five years project with industries to develop software for realizing highly reliable embedded systems. We have tackled reliability issues from the following aspects. (1)Design environment: we have developed a UML design verification tool that apply model checking techniques to improve the design quality of application software. (2)Operating environment: we have developed operating system supports that realize multiple execution of real-time operating systems and also developed operating systems with enhanced resource management. Both technologies contribute the realization of robust run-time environment. (3) Real-time environment: we have developed real-time garbage collection techniques for Jave. They prevent the suspension of applications that violates the correct behavior of real-time applications. Also, they reduce the effort of application programmers to avoid garbage-collection during important execution timing. We have obtained fruitful results from these three research themes, and some of them are actually used in industries. Furthermore, we have integrated the results to make synergetic effect of them. In order to demonstrate the effectiveness, we have conduct an experiment. In this paper, we introduce the project and its results.

In this position, we compare two promising approaches, model driven design and organic computing, and discuss a possible synergy between them.

In product line development (PLD), reusability is the key factor, and it is important to make architecture and components flexibly configurable. Thus far, several techniques for utilizing aspect-oriented technologies (AOTs) for PLD have been proposed. However, the application of AOTs to PLD is not simple and various issues related to the application, such as an invasive change problem that prevents reusability, have been reported. We have proposed an aspect oriented modeling (AOM) mechanism that separates concerns from their relationships with other concerns. In this study, we propose the application of our AOM to variability management and demonstrate the effectiveness of our approach with the help of a case study of embedded software.

In software design, it is important to encapsulate cross-cutting concerns, and the application of aspect-oriented technologies to design modeling is a significant challenge. We examined the design of software for embedded systems that exhibit complicated behavior and observed that aspect orientation is useful for designing such systems. In this paper, we propose aspect-oriented modeling for embedded software design. For this modeling, we introduce an aspect-oriented modeling mechanism based on state diagrams. We also demonstrate how the proposed modeling can be effectively applied to embedded software design based on an example.

It is important to design architecture to be steady throughout the evolution, or make the architecture to be the platform for a product family. In order to design software architecture to have such characteristics we have to analyze the commonality and differences among requirements on potential software that are supposed to be developed on the architecture, and then design software architecture so as to accommodate the commonality and differences. In this paper, we propose an approach for architectural design in which we analyze the requirements of potential software in terms of the impact on the architecture, considering multiple quality attributes. We perform a case study on an actual project that designed architecture for an on-board system for ITS systems to examine the usefulness of the technique. We also apply the technique to the same architectural design problem to demonstrate that it is applicable to the real problem.

It is difficult to design software to meet its goal on quality attributes, because there are many factors related to quality attributes, and the relationships between these factors and quality attributes are quite complicated. However, we do not have a systematic way to design software considering quality attributes. Consequently, we have many troubles in the attainment of required quality attributes in actual software development. We are examining a design method, aspect oriented design (AOD) based on the idea of "aspect orientedness" proposed in the programming community as aspect oriented programming. In AOD, aspects corresponding to quality attributes are considered separately, and software architectures suitable for each aspect are designed independently and woven into the final architecture. The authors introduce their approach and demonstrate the effectiveness of the approach using an example.