arXiv subjects

Jihwan Moon

Publications and source records attributed to Jihwan Moon.

6 recordsLinked to original sources

arXiv2026

Deep Reinforcement Learning for Adaptive Power Allocation in ISAC Systems with Mobile Target

In this paper, we study the power allocation for an integrated sensing and communication (ISAC) system which tracks a mobile target. We first model the problem as a Markov decision process, and then tackle it with a soft actor-critic (SAC) based deep reinforcement learning (DRL) approach. We also combine a Dirichlet policy, which naturally produces normalized continuous actions under random target motion. To exploit different features of sensing and communication operations, we carefully design a reward function such that the system can dynamically control power allocation to conserve resources. The simulation results demonstrate that the proposed scheme enhances tracking performance compared to other baselines while sustaining communication performance.

eess.SP↗

arXiv2026

Raon-Speech Technical Report

We present Raon-Speech, a top-performing 9B-parameter speech language model (SpeechLM) for English and Korean speech understanding, answering, and generation, and Raon-SpeechChat, a high-performing full-duplex extension for natural real-time conversation. Raon-Speech successfully transforms a pre-trained LLM into a SpeechLM that both understands and generates speech while preserving strong text capabilities. It trains on 1.38M hours of highly curated English and Korean speech and text datasets with the following training stages: (1) speech modules alignment, (2) end-to-end SpeechLM pre-training with knowledge distillation, and (3) multi-task preference optimization-based post-training. Across 42 English and Korean speech and text benchmarks, Raon-Speech establishes the strongest overall profile on speech-centric tasks in our comparison against eight similarly sized recent audio foundation models, including Qwen2.5-Omni and Fun-Audio-Chat, while preserving strong text question answering performance. Building upon it, Raon-SpeechChat enables natural full-duplex conversation by continual training on 119K hours of time-aligned real and synthetic dialogue data. It proceeds through three complementary training stages: (1) causal encoder adaptation, (2) full-duplex pre-training, (3) full-duplex fine-tuning for voice and role-control. On multiple full-duplex benchmarks, Raon-SpeechChat shows its clearest strengths on the turn-taking and interruption-sensitive behaviors covered by FDB v1.0, and remains competitive across the broader full-duplex evaluation suite. We open-source all model checkpoints, the training and inference pipeline, and an interactive demo.

cs.CL↗

arXiv2025

ReSpec: Relevance and Specificity Grounded Online Filtering for Learning on Video-Text Data Streams

The rapid growth of video-text data presents challenges in storage and computation during training. Online learning, which processes streaming data in real-time, offers a promising solution to these issues while also allowing swift adaptations in scenarios demanding real-time responsiveness. One strategy to enhance the efficiency and effectiveness of learning involves identifying and prioritizing data that enhances performance on target downstream tasks. We propose Relevance and Specificity-based online filtering framework (ReSpec) that selects data based on four criteria: (i) modality alignment for clean data, (ii) task relevance for target focused data, (iii) specificity for informative and detailed data, and (iv) efficiency for low-latency processing. Relevance is determined by the probabilistic alignment of incoming data with downstream tasks, while specificity employs the distance to a root embedding representing the least specific data as an efficient proxy for informativeness. By establishing reference points from target task data, ReSpec filters incoming data in real-time, eliminating the need for extensive storage and compute. Evaluating on large-scale datasets WebVid2M and VideoCC3M, ReSpec attains state-of-the-art performance on five zeroshot video retrieval tasks, using as little as 5% of the data while incurring minimal compute. The source code is available at https://github.com/cdjkim/ReSpec.

cs.CV↗

arXiv2025

Sample Selection via Contrastive Fragmentation for Noisy Label Regression

As with many other problems, real-world regression is plagued by the presence of noisy labels, an inevitable issue that demands our attention. Fortunately, much real-world data often exhibits an intrinsic property of continuously ordered correlations between labels and features, where data points with similar labels are also represented with closely related features. In response, we propose a novel approach named ConFrag, where we collectively model the regression data by transforming them into disjoint yet contrasting fragmentation pairs. This enables the training of more distinctive representations, enhancing the ability to select clean samples. Our ConFrag framework leverages a mixture of neighboring fragments to discern noisy labels through neighborhood agreement among expert feature extractors. We extensively perform experiments on six newly curated benchmark datasets of diverse domains, including age prediction, price prediction, and music production year estimation. We also introduce a metric called Error Residual Ratio (ERR) to better account for varying degrees of label noise. Our approach consistently outperforms fourteen state-of-the-art baselines, being robust against symmetric and random Gaussian label noise.

cs.LG↗

arXiv2022

A Codebook Design for FD-MIMO Systems with Multi-Panel Array

In this work, we study codebook designs for full-dimension multiple-input multiple-output (FD-MIMO) systems with a multi-panel array (MPA). We propose novel codebooks which allow precise beam structures for MPA FD-MIMO systems by investigating the physical properties and alignments of the panels. We specifically exploit the characteristic that a group of antennas in a vertical direction exhibit more correlation than those in a horizontal direction. This enables an economical use of feedback bits while constructing finer beams compared to conventional codebooks. The codebook is further improved by dynamically allocating the feedback bits on multiple parts such as beam amplitude and co-phasing coefficients using reinforcement learning. The numerical results confirm the effectiveness of the proposed approach in terms of both performance and computational complexity.

cs.IT↗

arXiv2019

Online Reinforcement Learning of X-Haul Content Delivery Mode in Fog Radio Access Networks

We consider a Fog Radio Access Network (F-RAN) with a Base Band Unit (BBU) in the cloud and multiple cache-enabled enhanced Remote Radio Heads (eRRHs). The system aims at delivering contents on demand with minimal average latency from a time-varying library of popular contents. Information about uncached requested files can be transferred from the cloud to the eRRHs by following either backhaul or fronthaul modes. The backhaul mode transfers fractions of the requested files, while the fronthaul mode transmits quantized baseband samples as in Cloud-RAN (C-RAN). The backhaul mode allows the caches of the eRRHs to be updated, which may lower future delivery latencies. In contrast, the fronthaul mode enables cooperative C-RAN transmissions that may reduce the current delivery latency. Taking into account the trade-off between current and future delivery performance, this paper proposes an adaptive selection method between the two delivery modes to minimize the long-term delivery latency. Assuming an unknown and time-varying popularity model, the method is based on model-free Reinforcement Learning (RL). Numerical results confirm the effectiveness of the proposed RL scheme.

eess.SP↗