Digital Library[ Search Result ]
A Large Language Model-based Multi-domain Recommender System using Model Merging
http://doi.org/10.5626/JOK.2025.52.6.548
Recent research in recommender systems has increasingly focused on leveraging pre-trained large language models (LLMs) to effectively understand the natural language information associated with recommendation items. While these LLM-based recommender systems achieve high accuracy, they have a limitation in that they require training separate recommendation models for each domain. This increases the costs of storing and inferring multiple models and makes it difficult to share knowledge across domains. To address this issue, we propose an LLM-based recommendation model that effectively operates across diverse recommendation domains by applying task vector-based model merging. During the merging process, knowledge distillation is utilized from individually trained domain-specific recommendation models to learn optimal merging weights. Experimental results show that our proposed method improves recommendation accuracy by an average of 2.75% across eight domains compared to recommender models utilizing existing model merging methods, while also demonstrating strong generalization performance in previously unseen domains.
An Inference Framework for Text-Based Sequential Recommendation Model Using Nearest Neighbor Mechanism
Junyoung Kim, Hyunsoo, Jongwuk Lee
http://doi.org/10.5626/JOK.2025.52.5.435
Sequential recommendation task aims to predict the next item to interact with based on users’ interaction history. Text-based recommendation models, which represent items as text, show improved performance in cold-start problems and zero-shot recommendation tasks. However, they suffer from textual bias and the lack of collaborative knowledge. To overcome these limitations, we propose a text-based recommendation model inference framework using the nearest neighbor mechanism. The proposed method leverages text-based recommendation models as a neighbor retriever model to search neighbors with similar preferences to the user and aggregate the neighbor information with existing recommendation results to improve recommendation performance. Experiments conducted on four datasets show that the proposed method consistently outperforms existing models, with performance improvement up to 25.27% on NDCG@50. Furthermore, the proposed method effectively complements collaborative knowledge and improves model explainability by providing recommendation rationale.
Recursive Compaction Method of LSM-tree based Key-value Store
Jongbin Kim, Seohui Son, Hyunsoo Cho, Hyungsoo Jung
http://doi.org/10.5626/JOK.2019.46.9.946
LSM-tree-based key-value stores exhibit an optimized structure for data writing operations and typically maintain the form of LSM tree by executing a compaction operation. The compaction operation which reads data from the storage device into memory for sorting it and writes back the result data in to the storage device several times causes some problems. In this paper, we analyzed the performance degradation and the write amplification caused by the compaction, and proposed a new compaction method known as recursive compaction. Recursive compaction alleviates the problems involving the compaction operation by utilizing multiple threads to perform multiple compactions at a time, handling read operation and garbage collection properly. We implemented this technique for Google LevelDB and analyzed the results.
Identifying Causes of an Accident in STPA Using the Scenario Table
http://doi.org/10.5626/JOK.2019.46.8.787
In recent years, the complexity of safety-critical systems has increased, along with the importance of the software. The software, which has become the control center of the safety system, generates control actions to control the system and then repeats the interaction of controls that re-enters the feedback generated. STPA (System Theoretic Process Analysis) is one of the hazard analysis techniques, and it analyzes the system from the perspective of the interaction of control then uses accident scenarios to identify and analyze the cause of unsafe control actions to derive safe requirements. In order to minimize omissions in the identification stage of STPA accident scenarios associated with safety requirements, in this paper we describe how to incorporate commonalities and complement vulnerabilities in the approaches described in previous studies. To do this, we propose the detailed procedure for identifying accident scenarios and the scenario table to assist them. The ultimately proposed scenario table is identified by applying it to the hazard analysis of the railway diorama system.
Rate Control Scheme for Improving Quality of Experience in the CoAP-based Streaming Environment
Hyunsoo Kang, Jiwoo Park, Kwangsue Chung
http://doi.org/10.5626/JOK.2017.44.12.1296
Recently, as the number of Internet of Things users has increased, IETF (Internet Engineering Task Force) has released the CoAP (Constrained Application Protocol). So Internet of Things have been researched actively. However, existing studies are difficult to adapt to streaming service due to low transmission rate that result from buffer underflow. In other words, one block is transmitted one block to client’s one request according to the internet environment of limited resources. The proposed scheme adaptively adjusts the rate of CON(Confirmable) message among all messages for predicting the exact network condition. Based on this, the number of blocks is determined by using buffer occupancy rate and content download rate. Therefore it improves the quality of user experience by mitigating playback interruption. Experimental results show that the proposed scheme solves the buffer underflow problem in Internet of Things streaming environment by controlling transmission rate according to the network condition.
Modified RTT Estimation Scheme for Improving Throughput of Delay-based TCP in Wireless Networks
Hyunsoo Kang, Jiwoo Park, Kwangsue Chung
In a wireless network, TCP causes the performance degradation because of mistaking packet loss, which is caused by characteristics of wireless link and throughput oscillation due to change of devices connected on a limited bandwidth. Delay based TCP is not affected by packet loss because it controls window size by using the RTT. Therefore, it can solve the problem of unnecessary degradation of the rate caused by misunderstanding reason of packet loss. In this paper, we propose an algorithm for improving the remaining problems by using delay based TCP. The proposed scheme can change throughput adaptively by adding the RTT, which rapidly reflects the network conditions to BaseRTT. It changes the weight of RTT and the increases and decreases window size based on the remaining amount of the buffer. The simulation indicated that proposed scheme can alleviate the throughput oscillation problem, as compared to the legacy TCP Vegas.
Data Block based User Authentication for Outsourced Data
Changhee Hahn, Hyunsoo Kown, Daeyeong Kim, Junbeom Hur
Recently, there has been an explosive increase in the volume of multimedia data that is available as a result of the development of multimedia technologies. More and more data is becoming available on a variety of web sites, and it has become increasingly cost prohibitive to have a single data server store and process multimedia files locally. Therefore, many service providers have been likely to outsource data to cloud storage to reduce costs. Such behavior raises one serious concern: how can data users be authenticated in a secure and efficient way? The most widely used password-based authentication methods suffer from numerous disadvantages in terms of security. Multi-factor authentication protocols based on a variety of communication channels, such as SMS, biometric, or hardware tokens, may improve security but inevitably reduce usability. To this end, we present a data block-based authentication scheme that is secure and guarantees usability in such a manner where users do nothing more than enter a password. In addition, the proposed scheme can be effectively used to revoke user rights. To the best of our knowledge, our scheme is the first data block-based authentication scheme for outsourced data that is proven to be secure without degradation in usability. An experiment was conducted using the Amazon EC2 cloud service, and the results show that the proposed scheme guarantees a nearly constant time for user authentication.
Improving Recall for Context-Sensitive Spelling Correction Rules using Conditional Probability Model with Dynamic Window Sizes
Hyunsoo Choi, Hyukchul Kwon, Aesun Yoon
The types of errors corrected by a Korean spelling and grammar checker can be classified into isolated-term spelling errors and context-sensitive spelling errors (CSSE). CSSEs are difficult to detect and to correct, since they are correct words when examined alone. Thus, they can be corrected only by considering the semantic and syntactic relations to their context. CSSEs, which are frequently made even by expert wiriters, significantly affect the reliability of spelling and grammar checkers. An existing Korean spelling and grammar checker developed by P University (KSGC 4.5) adopts hand-made correction rules for correcting CSSEs. The KSGC 4.5 is designed to obtain very high precision, which results in an extremely low recall. Our overall goal of previous works was to improve the recall without considerably lowering the precision, by generalizing CSSE correction rules that mainly depend on linguistic knowledge. A variety of rule-based methods has been proposed in previous works, and the best performance showed 95.19% of average precision and 37.56% of recall. This study thus proposes a statistics based method using a conditional probability model with dynamic window sizes. in order to further improve the recall. The proposed method obtained 97.23% of average precision and 50.50% of recall.
Security Enhanced Authentication Protocol in LTE With Preserving User Location Privacy
Changhee Hahn, Hyunsoo Kwon, Junbeom Hur
The number of subscribers in 4th generation mobile system has been increased rapidly. Along with that, preserving subscribers’ privacy has become a hot issue. To prevent users’ location from being revealed publicly is important more than ever. In this paper, we first show that the privacy-related problem exists in user authentication procedure in 4th generation mobile system, especially LTE. Then, we suggest an attack model which allows an adversary to trace a user, i.e. he has an ability to determine whether the user is in his observation area. Such collecting subscribers’ location by an unauthorized third party may yield severe privacy problem. To keep users’ privacy intact, we propose a modified authentication protocol in LTE. Our scheme has low computational overhead and strong secrecy so that both the security and efficiency are achieved. Finally, we prove that our scheme is secure by using the automatic verification tool ProVerif.
Search
Journal of KIISE
- ISSN : 2383-630X(Print)
- ISSN : 2383-6296(Electronic)
- KCI Accredited Journal
Editorial Office
- Tel. +82-2-588-9240
- Fax. +82-2-521-1352
- E-mail. chwoo@kiise.or.kr