In information retrieval, facet identification of a user query is an important task. If a search service can recognize the facets of a user’s query, it has the potential to offer users a much broader range of search results. Previous studies can enhance facet prediction by leveraging retrieved documents and related queries obtained through a search engine. However, there are challenges in extending it to other applications when a search engine operates as part of the model. First, search engines are constantly updated. Therefore, additional information may change during training and test, which may reduce performance. The second challenge is that public search engines cannot search for internal documents. Therefore, a separate search system needs to be built to incorporate documents from private domains within the company. We propose two strategies that focus on a framework that can predict facets by taking only queries as input without a search engine. The first strategy is multi-task learning to predict SERP. By leveraging SERP as a target instead of a source, the proposed model deeply understands queries without relying on external modules. The second strategy is to enhance the facets by combining Large Language Model (LLM) and the small model. Overall performance improves when small model and LLM are combined rather than facet generation individually.

Network slicing, a key technology of next-generation wireless networks, has undergone significant evolution from its inception as Dedicated Core Network (DCN) in 4G-LTE to its current state in 5G-Advanced. This paper provides a comprehensive analysis of network slicing enhancements across 3GPP releases 13 to 17, categorized into three phases: 5G-Basic (Release 15), early 5G-Evolution (Release 16), and advanced 5G-Evolution (Release 17). Furthermore, our study identifies persistent challenges in network slicing implementation and proposes innovative enhancements for 5G-Advanced (Release 18), including a novel machine learning-based approach to minimize service interruptions within a Registration Area (RA). This approach combines predictive insights from a Long Short-Term Memory (LSTM) model with a Dynamic Proportional Resource Allocation (DPRA) method for resource reconfiguration. Evaluation of the LSTM-DPRA scheme demonstrates significant performance improvements and reduced service interruptions compared to benchmark schemes, contributing to the development of more efficient and reliable network slicing.

Millimeter wave (mmWave) and terahertz (THz) communications have been considered as the key techniques to support extremely high data rates in the 6G system. One main limitation of the mmWave/THz communications is the severe path loss and low penetration power. For these reasons, it is expected that mmWave/THz communication will be mainly employed in the ultra-dense network (UDN) environment. In order to get the most out of the mmWave/THz UDN, a mobile should be associated to the base stations (BSs) providing a high quality-of-service (QoS). This task is challenging since the reliable path can be disappeared even with a small movement of a mobile. An aim of this paper is to propose a novel mobility management technique based on sensing and computer vision (CV). Our key idea is to predict the cell association from the visual sensing information and CV-based inference and decision. By extracting the geometric information of a mobile from the image and then using it for the downlink rate prediction, we preemptively switch the cell association in UDN. From the numerical evaluations on the realistic mmWave/THz UDN environments, we show that the proposed scheme achieves more than 30% throughput gain over the conventional mobility management techniques.

Recently, terahertz (THz) communication supported by the ultra-dense network (UDN) has received a great deal of attention as a means to satisfy stringent requirements in throughput, latency, and energy consumption in 6G. In the UDN supported by THz beamforming, handover, an action to change the base station (BS) serving the user, occurs frequently due to the small cell coverage and sudden line-of-sight (LoS) link blockage caused by the interruption of obstacles. To ensure the seamless connectivity in the THz UDN, mobility management, the process to identify the link deterioration and perform the handover, should be performed quickly and accurately. An aim of this paper is to propose a novel computer vision (CV)-aided framework to proactively control the mobility management in the THz communication regime. In our framework referred to as proactive computer vision-aided mobility management (P-CVMM), the position of the user (i.e., distance and angles) is extracted from the images via deep learning (DL)-based object detector and then exploited in predicting the optimal serving BS (S-BS). Since the sparse geometric channel parameters are immediately obtained without the complicated feedback process and the beam heading toward the user's future position can be generated without quantization, P-CVMM can predictively avoid radio link failure (RLF). Using the specially designed dataset called vision objects for mobility management (VOMM), we demonstrate that P-CVMM effectively avoids RLF and achieves more than 90% and 70% reduction in the localization error and handover interruption time (HIT) over the conventional schemes.

Millimeter-wave (mmWave) communications will play an important role in 5G and 6G communication systems as a means to support extremely high data rates. One main bottleneck of the mmWave communication is the severe signal attenuation caused by the foliage loss, atmospheric absorption, body and hand losses in the mmWave band. To compensate for the severe path loss, multiple-input-multiple-output (MIMO) antenna array-based beamforming has been widely used. Since the beams should be aligned with the signal propagation paths to get the most of beamforming gain, acquisition of accurate channel knowledge, i.e., channel estimation, is the key to the success of mmWave MIMO systems. In this paper, we propose a new type of deep learning (DL)-based parametric channel estimation technique. In our work, DL figures out the direct mapping between the received pilot signal and the sparse channel parameters characterizing the angular domain channel. By exploiting the long short-term memory (LSTM) as a main deep neural network (DNN) engine, we extract the temporally correlated features of time-varying channel parameters and make a fast yet accurate estimation with relatively small pilot overhead. From the numerical experiments, we show that the proposed scheme is effective in estimating the mmWave MIMO channel in various mmWave downlink environments.

Session-based recommendation (SR) predicts the next items from a sequence of previous items consumed by an anonymous user. Most existing SR models focus only on modeling intra-session characteristics but pay less attention to inter-session relationships of items, which has the potential to improve accuracy. Another critical aspect of recommender systems is computational efficiency and scalability, considering practical feasibility in commercial applications. To account for both accuracy and scalability, we propose a novel session-based recommendation with a random walk, namely S-Walk. Precisely, S-Walk effectively captures intra- and inter-session correlations by handling high-order relationships among items using random walks with restart (RWR). By adopting linear models with closed-form solutions for transition and teleportation matrices that constitute RWR, S-Walk is highly efficient and scalable. Extensive experiments demonstrate that S-Walk achieves comparable or state-of-the-art performance in various metrics on four benchmark datasets. Moreover, the model learned by S-Walk can be highly compressed without sacrificing accuracy, conducting two or more orders of magnitude faster inference than existing DNN-based models, making it suitable for large-scale commercial systems.

Session-based recommendation aims at predicting the next item given a sequence of previous items consumed in the session, e.g., on e-commerce or multimedia streaming services. Specifically, session data exhibits some unique characteristics, i.e., session consistency and sequential dependency over items within the session, repeated item consumption, and session timeliness. In this paper, we propose simple-yet-effective linear models for considering the holistic aspects of the sessions. The comprehensive nature of our models helps improve the quality of session-based recommendation. More importantly, it provides a generalized framework for reflecting different perspectives of session data. Furthermore, since our models can be solved by closed-form solutions, they are highly scalable. Experimental results demonstrate that the proposed linear models show competitive or state-of-the-art performance in various metrics on several real-world datasets.

Spectrum sensing is used to perceive the spectral environment over a wide frequency band. The multiple measurement vector (MMV) model can be applied to the spectrum sensing scenario since it enables jointly sparse signal recovery. In this paper, a novel spectrum sensing algorithm, referred to as multiple subspace matching pursuit (MSMP), is proposed to reduce the miss detection and false alarm events in the spectrum sensing. Numerical simulations demonstrate that the proposed algorithm shows the outstanding recovery performance with the reduction of the incorrect spectrum decisions.