Searcharxiv⌕ Search

arXiv subjects

Kijung Kim

Publications and source records attributed to Kijung Kim.

At least 19 recordsLinked to original sources
arXiv2026

Initiation of Interaction Detection Framework using a Nonverbal Cue for Human-Robot Interaction

This paper describes an initiation of interaction(IoI) detection framework without keywords for human-robot interaction(HRI) based on audio and vision sensor fusion in a domestic environment. In the proposed framework, the robot has its own audio and vision sensors, and can employ external vision sensor for stable human detection and tracking. When the user starts to speak while looking at the robot, the robot can localize his or her position by its sound source localization together with human tracking information. Then the robot can detect the IoI if it perceives the face of the speaker faces the robot. In case that the user does not speak directly, the robot can also detect the IoI if he or she looks at the robot for more than predefined periods of time. A state transition model for the proposed IoI detection framework is designed and verified by experiments with a mobile robot. In order to implement and associate our model in a robot architecture, all the components are implemented and integrated in the Robot Operating System(ROS) environment.

cs.CV↗
arXiv2025

Spectral-Adaptive Modulation Networks for Visual Perception

Recent studies have shown that 2D convolution and self-attention exhibit distinct spectral behaviors, and optimizing their spectral properties can enhance vision model performance. However, theoretical analyses remain limited in explaining why 2D convolution is more effective in high-pass filtering than self-attention and why larger kernels favor shape bias, akin to self-attention. In this paper, we employ graph spectral analysis to theoretically simulate and compare the frequency responses of 2D convolution and self-attention within a unified framework. Our results corroborate previous empirical findings and reveal that node connectivity, modulated by window size, is a key factor in shaping spectral functions. Leveraging this insight, we introduce a \textit{spectral-adaptive modulation} (SPAM) mixer, which processes visual features in a spectral-adaptive manner using multi-scale convolutional kernels and a spectral re-scaling mechanism to refine spectral components. Based on SPAM, we develop SPANetV2 as a novel vision backbone. Extensive experiments demonstrate that SPANetV2 outperforms state-of-the-art models across multiple vision tasks, including ImageNet-1K classification, COCO object detection, and ADE20K semantic segmentation.

cs.CV↗
arXiv2024

Point-Cloud Based Inverse Design of Free-Form Metamaterials Using Deep Generative Networks

Mechanical metamaterials enable precise control over structural properties, but their design method remains challenging due to their complex structure. Although additive manufacturing has expanded geometric freedom, navigating this vast and complex design space still requires computationally intensive simulations or expert-driven processes. Recently, artificial intelligence (AI)-driven design approaches have emerged to address these limitations, but many studies restrict their scope to parametric representations, limiting their generative capacity to predefined shapes. Here, we present a point cloud-based generative framework that enables the inverse design of 3D metamaterial without parametric constraints. Trained on a number of structurally valid unit cells, the present machine learning model learns geometric patterns, mitigates common connectivity issues inherent in point cloud generation. The proposed model constructs a latent space organized by mechanical properties and naturally clustered by unit cell types. By sampling this latent space, our method supports both property-guided inverse design and generation of topologically gradient transition between distinct unit cell types. This approach facilitates inverse design of 3D metamaterials with high geometric complexity.

cond-mat.soft↗
arXiv2023

SPANet: Frequency-balancing Token Mixer using Spectral Pooling Aggregation Modulation

Recent studies show that self-attentions behave like low-pass filters (as opposed to convolutions) and enhancing their high-pass filtering capability improves model performance. Contrary to this idea, we investigate existing convolution-based models with spectral analysis and observe that improving the low-pass filtering in convolution operations also leads to performance improvement. To account for this observation, we hypothesize that utilizing optimal token mixers that capture balanced representations of both high- and low-frequency components can enhance the performance of models. We verify this by decomposing visual features into the frequency domain and combining them in a balanced manner. To handle this, we replace the balancing problem with a mask filtering problem in the frequency domain. Then, we introduce a novel token-mixer named SPAM and leverage it to derive a MetaFormer model termed as SPANet. Experimental results show that the proposed method provides a way to achieve this balance, and the balanced representations of both high- and low-frequency components can improve the performance of models on multiple computer vision tasks. Our code is available at $\href{https://doranlyong.github.io/projects/spanet/}{\text{https://doranlyong.github.io/projects/spanet/}}$.

cs.CV↗
arXiv2021

Perfect Roman domination in middle graphs

The middle graph $M(G)$ of a graph $G$ is the graph obtained by subdividing each edge of $G$ exactly once and joining all these newly introduced vertices of adjacent edges of $G$. A perfect Roman dominating function on a graph $G$ is a function $f : V(G) \rightarrow \{0, 1, 2\}$ satisfying the condition that every vertex $v$ with $f(v)=0$ is adjacent to exactly one vertex $u$ for which $f(u)=2$. The weight of a perfect Roman dominating function $f$ is the sum of weights of vertices. The perfect Roman domination number is the minimum weight of a perfect Roman dominating function on $G$. In this paper, we give a characterization of middle graphs with equal Roman domination and perfect Roman domination numbers.

math.CO↗
arXiv2021

Restrained Italian domination in trees

Let $G=(V,E)$ be a graph. A subset $D$ of $V$ is a \textit{restrained dominating set} if every vertex in $V \setminus D$ is adjacent to a vertex in $D$ and to a vertex in $V \setminus D$. The \textit{restrained domination number}, denoted by $γ_r(G)$, is the smallest cardinality of a restrained dominating set of $G$. A function $f : V \rightarrow \{0, 1, 2\}$ is a \textit{restrained Italian dominating function} on $G$ if (i) for each vertex $v \in V$ for which $f(v)=0$, it holds that $\sum_{u \in N_G(v)} f(u) \geq 2$, (ii) the subgraph induced by $\{v \in V \mid f(v)=0 \}$ has no isolated vertices. The \textit{restrained Italian domination number}, denoted by $γ_{rI}(G)$, is the minimum weight taken over all restrained Italian dominating functions of $G$. It is known that $γ_r(G) \leq γ_{rI}(G) \leq 2γ_r(G)$ for any graph $G$. In this paper, we characterize the trees $T$ for which $γ_r(T) = γ_{rI}(T)$, and we also characterize the trees $T$ for which $γ_{rI}(T) = 2γ_r(T)$.

math.CO↗
arXiv2020

On $k$-rainbow domination in middle graphs

Let $G$ be a finite simple graph with vertex set $V(G)$ and edge set $E(G)$. A function $f : V(G) \rightarrow \mathcal{P}(\{1, 2, \dotsc, k\})$ is a \textit{$k$-rainbow dominating function} on $G$ if for each vertex $v \in V(G)$ for which $f(v)= \emptyset$, it holds that $\bigcup_{u \in N(v)}f(u) = \{1, 2, \dotsc, k\}$. The weight of a $k$-rainbow dominating function is the value $\sum_{v \in V(G)}|f(v)|$. The \textit{$k$-rainbow domination number} $γ_{rk}(G)$ is the minimum weight of a $k$-rainbow dominating function on $G$. In this paper, we initiate the study of $k$-rainbow domination numbers in middle graphs. We define the concept of a middle $k$-rainbow dominating function, obtain some bounds related to it and determine the middle $3$-rainbow domination number of some classes of graphs. We also provide upper and lower bounds for the middle $3$-rainbow domination number of trees in terms of the matching number. In addition, we determine the $3$-rainbow domatic number for the middle graph of paths and cycles.

cs.DM↗
arXiv2020

The Italian bondage and reinforcement numbers of digraphs

An \textit{Italian dominating function} on a digraph $D$ with vertex set $V(D)$ is defined as a function $f : V(D) \rightarrow \{0, 1, 2\}$ such that every vertex $v \in V(D)$ with $f(v) = 0$ has at least two in-neighbors assigned $1$ under $f$ or one in-neighbor $w$ with $f(w) = 2$. The \textit{weight} of an Italian dominating function $f$ is the value $ω(f) = f(V(D)) = \sum_{u \in V(D)} f(u)$. The \textit{Italian domination number} of a digraph $D$, denoted by $γ_I(D)$, is the minimum taken over the weights of all Italian dominating functions on $D$. The \textit{Italian bondage number} of a digraph $D$, denoted by $b_I(D)$, is the minimum number of arcs of $A(D)$ whose removal in $D$ results in a digraph $D'$ with $γ_I(D') > γ_I(D)$. The \textit{Italian reinforcement number} of a digraph $D$, denoted by $r_I(D)$, is the minimum number of extra arcs whose addition to $D$ results in a digraph $D'$ with $γ_I(D') < γ_I(D)$. In this paper, we initiate the study of Italian bondage and reinforcement numbers in digraphs and present some bounds for $b_I(D)$ and $r_I(D)$. We also determine the Italian bondage and reinforcement numbers of some classes of digraphs.

cs.DM↗
arXiv2020

The Italian domination numbers of some products of directed cycles

An Italian dominating function on a digraph $D$ with vertex set $V(D)$ is defined as a function $f : V(D) \rightarrow \{0, 1, 2\}$ such that every vertex $v \in V(D)$ with $f(v) = 0$ has at least two in-neighbors assigned $1$ under $f$ or one in-neighbor $w$ with $f(w) = 2$. In this paper, we determine the exact values of the Italian domination numbers of some products of directed cycles.

cs.DM↗
arXiv2018

Two-valenced association schemes and the Desargues theorem

The main goal of the paper is to establish a sufficient condition for a two-valenced association scheme to be schurian and separable. To this end, an analog of the Desargues theorem is introduced for a noncommutative geometry defined by the scheme in question. It turns out that if the geometry has enough many Desarguesian configurations, then under a technical condition the scheme is schurian and separable. This result enables us to give short proofs for known statements on the schurity and separability of quasi-thin and pseudocyclic schemes. Moreover, by the same technique we prove a new result: given a prime $p$, any $\{1,p\}$-scheme with thin residue isomorphic to an elementary abelian $p$-group of rank greater than two, is schurian and separable.

math.CO↗
arXiv2018

Schurity and separability of quasiregular coherent configurations

A permutation group is said to be quasiregular if every its transitive constituent is regular, and a quasiregular coherent configuration can be thought as a combinatorial analog of such a group: the transitive constituents are replaced by the homogeneous components. In this paper, we are interested in the question when the configuration is schurian, i.e., formed by the orbitals of a permutation group, or/and separable, i.e., uniquely determined by the intersection numbers. In these terms, an old result of Hanna Neumann is, in a sense, dual to the statement that the quasiregular coherent configurations with cyclic homogeneous components are schurian. In the present paper, we (a) establish the duality in a precise form and (b) generalize the latter result by proving that a quasiregular coherent configuration is schurian and separable if the groups associated with homogeneous components have distributive lattices of normal subgroups.

math.CO↗
arXiv2016

On commutative $p$-schemes of order $p^4$

In this article, we consider the existence and schurity problem on commutative $p$-schemes of order $p^4$. Using the thin radical and thin residue, we give sufficient conditions for such $p$-schemes to be schurian. We also give questions related to our results.

math.GR↗
arXiv2015

A family of non-Schurian $p$-Schur rings over groups of order $p^3$

Recently, it was proved that every commutative $p$-Schur ring over a group of order $p^3$ is Schurian. In this article, we consider the Schurity problem of non-commutative $p$-Schur rings over groups of order $p^3$. In particular, it is given a family of non-Schurian $p$-Schur rings over groups of order $p^3$.

math.RA↗
arXiv2015

Isomorphism classes of association schemes induced by Hadamard matrices

Every Hadamard matrix $H$ of order $n > 1$ induces a graph with $4n$ vertices, called the Hadamard graph $Γ(H)$ of $H$. Since $Γ(H)$ is a distance-regular graph with diameter $4$, it induces a $4$-class association scheme $(Ω, S)$ of order $4n$. In this article we deal with fission schemes of $(Ω, S)$ under certain conditions, and for such a fission scheme we estimate the number of isomorphism classes with the same intersection numbers as the fission scheme.

math.CO↗
arXiv2015

Terwilliger algebras of wreath products by 3-equivalent schemes

Recently G. Bhattacharyya, S.Y. Song and R. Tanaka began to study Terwilliger algebras of wreath products of one-class association schemes. K. Kim determined the structure of Terwilliger algebras of wreath products by one-class association schemes or quasi-thin schemes. In this paper, we study Terwilliger algebras of wreath products by $3$-equivalenced schemes.

math.RA↗
arXiv2012

Terwilliger algebras of wreath products by quasi-thin schemes

The structure of Terwilliger algebras of wreath products by thin schemes or one-class schemes was studied in [A. Hanaki, K. Kim, Y. Maekawa, Terwilliger algebras of direct and wreath products of association schemes, J. Algebra 343 (2011) 195--200]. In this paper, we will consider the structure of Terwilliger algebras of wreath products by quasi-thin schemes. This gives a generalization of their result.

math.RT↗