• Journal of Information Processing Systems
• /
• v.19 no.6
• /
• pp.803-816
• /
• 2023

Explainable artificial intelligence is a method that explains how a complex model (e.g., a deep neural network) yields its output from a given input. Recently, graph-type data have been widely used in various fields, and diverse graph neural networks (GNNs) have been developed for graph-type data. However, methods to explain the behavior of GNNs have not been studied much, and only a limited understanding of GNNs is currently available. Therefore, in this paper, we propose an explanation method for node classification using graph convolutional networks (GCNs), which is a representative type of GNN. The proposed method finds out which features of each node have the greatest influence on the classification of that node using GCN. The proposed method identifies influential features by backtracking the layers of the GCN from the output layer to the input layer using the gradients. The experimental results on both synthetic and real datasets demonstrate that the proposed explanation method accurately identifies the features of each node that have the greatest influence on its classification.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.21 no.9
• /
• pp.1414-1421
• /
• 1997

This paper examines the physics and mechanics governing the dynamic interaction between physical systems and suggests the four structures of bond graph prototypes, considered as a general model that can promise their dynamic behavior physically resonable. The bond graph prototypes originating from the paper are more realistic junction structures than those used to model dynamic systems conventionally by bond graph standards in whether physical constraints are involved or not when the energy exchange between two dynamic components arises. It is shown that the bond graph prototypes are dynamic or energetic in their describing equations compared to the bond graph standards, and connectivity and causality are properties of dynamic systems upon which the steps developed in this paper for the bond graph prototypes are wholly based and their definitions an concepts are highly emphasized all through the paper.

• Journal of Applied and Pure Mathematics
• /
• v.5 no.5_6
• /
• pp.363-373
• /
• 2023

In this paper we investigate the 4-total mean cordial labeling behavior of arrow graphs, shell-Butterfly graph and graphs obtained by joining two copies of shell graphs by a path.

• Nuclear Engineering and Technology
• /
• v.29 no.6
• /
• pp.433-443
• /
• 1997

In this work, an analytic approach to the dependability of software in the operational phase is suggested with special attention to the hardware fault effects on the software behavior : The hardware faults considered are memory faults and the dependability measure in question is the reliability. The model is based on the simple reliability theory and the graph theory which represents the software with graph composed of nodes and arcs. Through proper transformation, the graph can be reduced to a simple two-node graph and the software reliability is derived from this graph. Using this model, we predict the reliability of an application software in the digital system (ILS) in the nuclear power plant and show the sensitivity of the software reliability to the major physical parameters which affect the software failure in the normal operation phase. We also found that the effects of the hardware faults on the software failure should be considered for predicting the software dependability accurately in operation phase, especially for the software which is executed frequently. This modeling method is particularly attractive for the medium size programs such as the microprocessor-based nuclear safety logic program.

• The Pure and Applied Mathematics
• /
• v.12 no.1
• /
• pp.47-55
• /
• 2005

For topological spaces X, Y and the function f : X → Y, it induces a function gr(f) : X → X x Y defined as gr(f)(χ) = (χ, f(χ)), for every χ ∈ X. It deals with some preliminary investigations relating to the behavior of functions and its graph functions. It has also been found that continuous functions are homotopic if and only if their graph functions are homotopic.

• The KIPS Transactions:PartC
• /
• v.13C no.1 s.104
• /
• pp.45-54
• /
• 2006

Nowadays, a lot of techniques have been applied for the detection of malicious behavior. However, the current techniques taken into practice are facing with the challenge of much variations of the original malicious behavior, and it is impossible to respond the new forms of behavior appropriately and timely. There are also some limitations can not be solved, such as the error affirmation (positive false) and mistaken obliquity (negative false). With the questions above, we suggest a new method here to improve the current situation. To detect the malicious code, we put forward dealing with the basic source code units through the conceptual graph. Basically, we use conceptual graph to define malicious behavior, and then we are able to compare the similarity relations of the malicious behavior by testing the formalized values which generated by the predefined graphs in the code. In this paper, we show how to make a conceptual graph and propose an efficient method for similarity measure to discern the malicious behavior. As a result of our experiment, we can get more efficient detection rate.

• Journal of Applied and Pure Mathematics
• /
• v.6 no.1_2
• /
• pp.55-69
• /
• 2024

Let a graph G = (V, E) be a (p, q) graph. Define $${\rho}=\{\array{{\frac{p}{2}} && p\;\text{is even} \\ {\frac{p-1}{2}} && p\;\text{is odd,}}$$ and M = {±1, ±2, … ± 𝜌} called the set of labels. Consider a mapping λ : V → M by assigning different labels in M to the different elements of V when p is even and different labels in M to p - 1 elements of V and repeating a label for the remaining one vertex when p is odd. The labeling as defined above is said to be a pair mean cordial labeling if for each edge uv of G, there exists a labeling $\frac{{\lambda}(u)+{\lambda}(v)}{2}$ if λ(u) + λ(v) is even and $\frac{{\lambda}(u)+{\lambda}(v)+1}{2}$ if λ(u) + λ(v) is odd such that ${\mid}\bar{\mathbb{s}}_{{\lambda}_1}-\bar{\mathbb{s}}_{{\lambda}^c_1}{\mid}\,{\leq}\,1$ where $\bar{\mathbb{s}}_{{\lambda}_1}$ and $\bar{\mathbb{s}}_{{\lambda}^c_1}$ respectively denote the number of edges labeled with 1 and the number of edges not labeled with 1. A graph G with a pair mean cordial labeling is called a pair mean cordial graph. In this paper, we investigate the pair mean cordial labeling behavior of some union of graphs.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.17 no.2
• /
• pp.87-97
• /
• 2018

Bus, a major transportation mode, doesn't have a systematical evaluation system for dangerous driving behavior yet. This paper analyzes the characteristics and pattern of bus driving behavior using Digital Tacho Graph(DTG) data on express bus. 8 types of dangerous driving behavior were considered according to timeslot, the day of week and weather condition. As results, rapid acceleration, rapid left right turn and rapid deceleratio accounted for more than 97% and relatively high percentages were shown in dawn, on Friday and on the clear day, respectively. From the statistical analysis, correlation between the dangerous driving types and difference according the timeslot were found, and 3 groups considering the level of the dangerous driving were suggested. This study contributes to setting an efficient and reliable eduction system for using driving simulators.

• Journal of Convergence for Information Technology
• /
• v.11 no.4
• /
• pp.55-63
• /
• 2021

Recently, the incidence rate of malicious codes is over tens of thousands of cases, and it is known that it is almost impossible to detect/respond all of them. This study proposes a method for detecting multiple behavior patterns based on a graph database as a new method for dealing with malicious codes. Traditional dynamic analysis techniques and has applied a method to design and analyze graphs of representative associations malware pattern(process, PE, registry, etc.), another new graph model. As a result of the pattern verification, it was confirmed that the behavior of the basic malicious pattern was detected and the variant attack behavior(at least 5 steps), which was difficult to analyze in the past. In addition, as a result of the performance analysis, it was confirmed that the performance was improved by about 9.84 times or more compared to the relational database for complex patterns of 5 or more steps.

• International Journal of Contents
• /
• v.10 no.3
• /
• pp.41-46
• /
• 2014

A cause-effect graph considers only the desired external behavior of a system by identifying input-output parameter relationships in the specification. When testing a software system with cause-effect graphs, it is important to derive a moderate number of tests while avoiding loss in fault detection ability. Pairwise testing is known to be effective in determining errors while considering only a small portion of the input space. In this paper, we present a new testing technique that generates pairwise tests from a cause-effect graph. We use a Boolean Satisbiability (SAT) solver to generate pairwise tests from a cause-effect graph. The Alloy language is used for encoding the cause-effect graphs and its SAT solver is applied to generate the pairwise tests. Using a SAT solver allows us to effectively manage constraints over the input parameters and facilitates the generation of pairwise tests, even in the situations where other techniques fail to satisfy full pairwise coverage.