Presentation Details

10:00~10:50     [교수발표]

• Recent Progress on Generative Search and Recommendation 

• 이종욱 교수

• As the volume of data continues to grow, search and recommendation have become essential for meeting users' information needs. Specifically, generative search and recommendation have introduced a new methodology that reframes traditional problems, such as query-document matching in search and user history-preference matching in recommendation. In this talk, we will offer a comprehensive review of the current research on generative search and recommendation, organizing it under a unified framework. We will also highlight the challenges and suggest future research directions. 

• Compact 3D Gaussian Representation for Radiance Field 

• 박은병 교수

• Neural Radiance Fields (NeRFs) have demonstrated remarkable potential in capturing complex 3D scenes with high fidelity. However, one persistent challenge that hinders the widespread adoption of NeRFs is the computational bottleneck due to the volumetric rendering. On the other hand, 3D Gaussian splatting (3DGS) has recently emerged as an alternative representation that leverages a 3D Gaussisan-based representation and adopts the rasterization pipeline to render the images rather than volumetric rendering, achieving very fast rendering speed and promising image quality. However, a significant drawback arises as 3DGS entails a substantial number of 3D Gaussians to maintain the high fidelity of the rendered images, which requires a large amount of memory and storage. To address this critical issue, we place a specific emphasis on two key objectives: reducing the number of Gaussian points without sacrificing performance and compressing the Gaussian attributes, such as view-dependent color and covariance. To this end, we propose a learnable mask strategy that significantly reduces the number of Gaussians while preserving high performance. In addition, we propose a compact but effective representation of view-dependent color by employing a grid-based neural field rather than relying on spherical harmonics. Finally, we learn codebooks to compactly represent the geometric attributes of Gaussian by vector quantization. With model compression techniques such as quantization and entropy coding, we consistently show over 25× reduced storage and enhanced rendering speed, while maintaining the quality of the scene representation, compared to 3DGS. Our work provides a comprehensive framework for 3D scene representation, achieving high performance, fast training, compactness, and real-time rendering. 

11:00~12:00     [초청강연]

• Generative Modeling for Photorealistic 3D Digital Humans 

• 주한별 교수 (서울대학교)

• Abstract: In this talk, I will present our latest research on developing generative models for creating highly realistic 3D digital humans. Three state-of-the-art approaches will be introduced: NCHO (ICCV 2023) for learning neural 3D composition of humans and clothing, Chupa (ICCV 2023) for creating 3D clothed humans using 2D diffusion probabilistic models, and GALA (CVPR 2024) for generating animatable layered assets from a single 3D scan. Key challenges, methodologies, and results will be discussed, along with insights into broader impacts and future directions of this field. 

• Biography: Hanbyul Joo is an assistant professor at Seoul National University (SNU) in the Department of Computer Science and Engineering. Before joining SNU, Hanbyul was a Research Scientist at Facebook AI Research (FAIR), Menlo Park. Hanbyul received his PhD from the Robotics Institute at Carnegie Mellon University, Hanbyul is a recipient of the Samsung Scholarship and the Best Student Paper Award in CVPR 2018. 

13:30~14:20     [교수발표]

• Recent  Progress on Deepfake Detection and Generation Research  

• 우사이먼성일 교수

• Deepfakes have become a critical social problem, and detecting them is of utmost importance. Detecting high-quality deepfake videos from widely released datasets are more straightforward to detect than low-quality ones. Most of the prior research achieve above 90% accuracy for detecting the high-quality deepfake videos from the open dataset. However, in real life, many deepfake videos that are leaked through social networks such as YouTube and instant messaging applications are highly compressed. As a result, the distributed video's resolution becomes extremely lower, making highly accurate detection methods harder. In this work, we present the current status, several challenges, and possible solutions to improve detection of different types of deepfakes. 

• Self-Supervised Learning and Its Applications 

• 이한국 교수

• Self-supervised learning, which learns by constructing artificial labels from only the input signals, has recently gained considerable attention for learning semantic representations from unlabeled data. In this talk, I will first introduce recent self-supervised learning techniques such as contrastive learning, clustering, and masked autoencoders. I will then present my recent work that incorporates these techniques into various applications, including few-shot learning, generative modeling, molecular and tabular representation learning. 

14:30~15:20     [교수발표]

• Providing Explanations for Unsupervised Deep Learning Models 

• 박호건 교수

• Node representation learning, such as Graph Neural Networks (GNNs), has emerged as a pivotal method in machine learning. The demand for reliable explanation generation surges, yet unsupervised models remain underexplored in this regard. To bridge this gap, we introduce a method for generating counterfactual (CF) explanations in unsupervised node representation learning. We identify the most important subgraphs that cause a significant change in the k-nearest neighbors of a node of interest in the learned embedding space upon perturbation. The k-nearest neighbor-based CF explanation method provides simple, yet pivotal, information for understanding unsupervised downstream tasks, such as top-k link prediction and clustering. Consequently, we introduce UNR-Explainer for generating expressive CF explanations for Unsupervised Node Representation learning methods based on a Monte Carlo Tree Search (MCTS). The proposed method demonstrates superior performance on diverse datasets for unsupervised GraphSAGE and DGI. Lastly, we introduce our recent efforts for unsupervised generative models. 

• PEMA: An Offsite-Tunable Plug-in External Memory Adaptation for Language Models 

• 박진영 교수

• Making neural networks remember over the long term has been a longstanding issue. Although several external memory techniques have been introduced, most focus on retaining recent information in the short term. Regardless of its importance, information tends to be fatefully forgotten over time. In this talk, I will present Memoria, a memory system for artificial neural networks, drawing inspiration from humans and applying various neuroscientific and psychological theories. The experimental results prove the effectiveness of Memoria in the diverse tasks of sorting, language modeling, and classification, surpassing conventional techniques. Engram analysis reveals that Memoria exhibits the primacy, recency, and temporal contiguity effects which are characteristics of human memory. 

15:30~16:30     [초청강연]

• Online Trust & Safety Series: Countering Online Hate Speech with AI 

• Prof. Roy Ka-Wei Lee (Singapore University of Technology and Design) 

• Abstract: As hate speech proliferates across digital platforms, the challenge of identifying and mitigating harmful content has become a critical issue for online trust and safety. In this talk, we explore how artificial intelligence (AI) is being leveraged to counter online hate speech effectively. We will examine the latest advancements in AI-driven content moderation, with a particular focus on detecting nuanced and context-dependent forms of hate, such as in memes and multilingual content. By addressing ethical challenges, including biases in AI models and the need for cultural sensitivity, this session will provide insights into how AI can balance the fine line between protecting free speech and fostering safer online communities. 

• Biography: Roy Ka-Wei Lee is an Assistant Professor at the Information Systems Technology and Design Pillar, Singapore University of Technology and Design. His research focuses on the intersection of data mining, computational social science, social computing, and natural language processing. Roy's particular interest lies in understanding user behaviours across multiple social networks and promoting online safety. Because of his efforts in designing solutions that combat online hate speech and misinformation, Roy has been appointed Adjunct Senior Scientist at Singapore's Centre for Advanced Technologies in Online Safety (CATOS). Roy is currently leading the Social AI Studio, a dedicated research group that strives to develop cutting-edge social artificial intelligence systems. His work has been recognized and published in renowned conferences and journals. Roy also actively contributes to the academic community by serving on program committees and acting as a reviewer for these esteemed conferences and journals. Roy is an IEEE Senior Member. 

16:30~18:00     [포스터 세션]

• Poster Session Details 참조