There were fierce debates on whether the non-linear embedding propagation of GCNs is appropriate to GCN-based recommender systems. It was recently found that linear shows better accuracy than propagation. Since this phenomenon discovered especially in systems, it required we carefully analyze linearity and non-linearity issue. In work, therefore, revisit issues i) which or ii) factors users/items decide linearity/non-linearity We propose a novel Hybrid method collaborative filtering (HMLET,...

In this paper, we focus on multimedia recommender systems using graph convolutional networks (GCNs) where the multimodal features as well user-item interactions are employed together. Our study aims to exploit more effectively in order accurately capture users' preferences for items. To end, point out following two limitations of existing GCN-based systems: (L1) although interacted items by a user can reveal her items, methods utilize GCN designed only capturing collaborative signals,...

We focus on the medication recommendation problem aiming to recommend accurate medications for a patient’s current visit. Most existing methods this utilize health status, prescribed at her past visits, and an Electronic Health Records (EHR) graph which represents whether have been co-prescribed. However, we point out their two limitations: (1) they difficulty in utilizing only status similar regardless of are visits or other patients’ visits; (2) that ever co-prescribed, EHR does not...

We address the multimedia recommendation problem, which utilizes items' multimodal features, such as visual and textual modalities, in addition to interaction information. While a number of existing recommender systems have been developed for this we point out that none these methods individually capture influence each modality at level. More importantly, experimentally observe learning procedures works fail preserve intrinsic modality-specific properties items. To above limitations, propose...

We address the medication recommendation problem, which aims to recommend effective medications for a patient's current visit by utilizing information (e.g., diagnoses and procedures) given at past visits. While there exist number of recommender systems designed this we point out that they are challenged in accurately capturing relation (spec., degree relevance) between each visits patient when obtaining her health status, is basis recommending medications. To limitation, propose novel...

Many cyberattacks start with disseminating phishing URLs. When clicking these URLs, the victim's private information is leaked to attacker. There have been proposed several machine learning methods detect However, it still remains under-explored URLs evasion, i.e., that pretend be benign by manipulating patterns. In many cases, attacker i) reuses prepared web pages because making a completely brand-new set costs non-trivial expenses, ii) prefers hosting companies do not require and are...

As our lives become increasingly dependent on computer software, the threat of malware attacks is getting greater. By slightly modifying previous version to avoid detection, attackers can continuously release new malwares with ease. However, released by a group authors might contain some evidence among them that they are developed same authors. Such information be used for digital forensics, law enforcement, and deeper analysis malwares. In this paper, we propose graph-based approach...

Multi-agent trajectory prediction is crucial to autonomous driving and understanding the surrounding environment. Learning-based approaches for multi-agent prediction, such as primarily relying on graph neural networks, transformers, hypergraph have demonstrated outstanding performance real-world datasets in recent years. However, transformer-based method yet be explored. Therefore, we present a MultiscAle Relational Transformer (MART) network prediction. MART transformer architecture...

This paper addresses the problem of multimedia recommendation that additionally utilizes data, such as visual and textual modalities items along with user-item interaction information. Existing recommender systems assume all non-interacted a user have same degree negativity, thus regarding them candidates for negative samples when training model. However, this claims user?s do not negativity. We classify these into two kinds different characteristics: unknown uninteresting items. Then, we...

피싱을 통한 사이버 범죄가 늘어나고 있다. 피싱이 초래하는 피해를 방지하기 위해 콘텐츠 기반, URL 문자열 기반 등 많은 피싱 관련 연구들이 진행되어 왔다. 방법은 웹 페이지 콘텐츠를 다운로드하고 분석하는 방법으로, 보안상 위험이 따르는 단점이 존재한다. 패턴을 분석하고 이를 탐지에 사용한다. 본 논문에서는 기존 연구로부터 확인된 URL의 경향에서 착안하여, 문자열을 그래프로 구축하고 Random Walk with Restart, Belief Propagation과 같은 그래프 추론 알고리즘과 DeepWalk, Node2vec과 임베딩 기법을 통해 여부를 예측한다. 우리의 탐지 방법과 분류 알고리즘을 활용한 방법을 비교한 결과, 방법이 모든 정확도 척도에서 더 높은 성능을 보였다.

Many cyberattacks start with disseminating phishing URLs. When clicking these URLs, the victim's private information is leaked to attacker. There have been proposed several machine learning methods detect However, it still remains under-explored URLs evasion, i.e., that pretend be benign by manipulating patterns. In many cases, attacker i) reuses prepared web pages because making a completely brand-new set costs non-trivial expenses, ii) prefers hosting companies do not require and are...

In this paper, we focus on multimedia recommender systems using graph convolutional networks (GCNs) where the multimodal features as well user-item interactions are employed together. Our study aims to exploit more effectively in order accurately capture users' preferences for items. To end, point out following two limitations of existing GCN-based systems: (L1) although interacted items by a user can reveal her items, methods utilize GCN designed only capturing collaborative signals,...

We address the medication recommendation problem, which aims to recommend effective medications for a patient's current visit by utilizing information (e.g., diagnoses and procedures) given at past visits. While there exist number of recommender systems designed this we point out that they are challenged in accurately capturing relation (spec., degree relevance) between each visits patient when obtaining her health status, is basis recommending medications. To limitation, propose novel...

This study was conducted to find out the characteristics of Meal-kit products that have recently grown rapidly in Korea. To this end, Topic Modeling and PageRank analysis were on online news articles. It is an easy identify main topics surrounding using Modeling, shows influence relationship well, so it various product. As a result study, delivery, online, cooking can be cited as products. used useful data for corporate marketing also predict new industries.

For autonomous driving, the ability to predict future from past trajectory is crucial. In this study, a graphical Y-branched multi-tasking network (GYM-Net) proposed. The GYM-Net predicts multi-scale features using multi-task learning. both final position after ten-time steps and based on previous of five-time steps. Using NBA dataset that widely used agents are active in multi-agent predictions, we assessed performance our model. model 2.37 m, 2.31 3.87 3.81 m for evaluation metrics avgADE,...

There were fierce debates on whether the non-linear embedding propagation of GCNs is appropriate to GCN-based recommender systems. It was recently found that linear shows better accuracy than propagation. Since this phenomenon discovered especially in systems, it required we carefully analyze linearity and non-linearity issue. In work, therefore, revisit issues i) which or ii) factors users/items decide linearity/non-linearity We propose a novel Hybrid Method Linear non-linEar collaborative...