Loading...

학술행사

Korean AI Association

  >   학술행사   >   국내학술대회

국내학술대회

튜토리얼
 
유재준 교수 (UNIST)
 
Title: Recent trends of text-to-image translation models
 

Abs

Image synthesis has been a hot topic in computer vision and graphics, showing remarkable performance in various applications such as image translation, image editing, etc. As research on generative modeling has matured, it has become relatively easy to create high-resolution photorealistic images. However, today's users want models that give them full control over the creation process beyond simply generating high-quality images. One intuitive way to let the user easily control the generative model is to use a simple text input as a condition to generate the desired image or to manipulate specific objects in the image. In this talk, I will present recent trends in the development of text-to-image translation models from GANs to diffusion models. 

 

Bio

I am an Assistant Professor and Director of the Laboratory of Advanced Imaging Technology (LAIT) in Graduate School of Artificial Intelligence at Ulsan National Institute of Science and Technology (UNIST). I am also an Affiliated Professor in Electrical Engineering Department. My main research area lies at the intersection of computer vision, signal processing, and machine learning. I have a strong interest in generative models, representation learning, and the use of signal processing theories for solving various inverse problems from low-level computer vision (e.g., natural image recovery) to medical imaging tasks (e.g., image reconstruction of CT, MRI, Microscopy, etc.). I also enjoy making small puns (What is "LAIT"?). Prior to joining UNIST, from 2019 to 2021, I was a postdoctoral researcher with Michael Unser at Biomedical Imaging Group, EPFL, Switzerland. From 2018 to 2019, I spent two years as an AI research scientist at NAVER Clova AI. Before then, I received my B.S., M.S., and Ph.D. degrees from KAIST under the supervision of Jong Chul Ye.
 
 
주재걸 교수 (KAIST)
 
Title: How A Diffusion Model Works in Text-to-Image Translation
 

Abs

Recently, we have seen remarkable progress in image generation and translation. In particular, Text-to-Image Translation, which synthesizes high-quality images reflecting the semantic meanings of an input text. Diffusion models are play a major role in making such a significant progress. In this talk, I will present how diffusion models work in detail and discuss the future research directions. 

 

Bio

Jaegul Choo is currently an associate professor in the Graduate School of Artificial Intelligence at KAIST. He has been an assistant professor in the Dept. of Computer Science and Engineering at Korea University from 2015 to 2019 and then an associate professor in the Dept. of Artificial Intelligence at Korea University in 2019. He received M.S in the School of Electrical and Computer Engineering at Georgia Tech in 2009 and Ph.D in the School of Computational Science and Engineering at Georgia Tech in 2013, advised by Prof. Haesun Park. From 2011 to 2014, he has been a research scientist at Georgia Tech. He earned his B.S in the Dept. of Electrical and Computer Engineering at Seoul National University.
 
 
 
김성웅 박사 (Kakao Brain)
 
Title: Large-Scale Agent Learning
 

Abs

For realizing artificial general intelligence (AGI), there are two necessary features that an AI system has to possess: task generalization and self-learning. While a number of recent learning paradigms such as meta-learning, automated learning, continual learning, and generative pre-training aim to achieve these features, agent learning is also one of them and most of all is closely in line with how a human perform multiple tasks through trial-and-error learning. In particular, agent learning can be defined as learning from experiences to perform optimal actions given observations and involves reinforcement learning as a core component. On top of reinforcement learning, it generally combines sequential modeling and world modeling.

This tutorial will first review the basic reinforcement learning briefly and then focus on recent distributed deep reinforcement learning for large-scale agent learning. Throughout the tutorial, a number of representative agent learning problems including sparse rewards, high-dimensional state space, procedurally generated and partially observable environments will also be introduced and how these problems can be solved using recent machine learning algorithms will be described. In addition, recent approaches for generalizable agent learning based on multi-task / multi-modal learning, self-supervised representation learning, world modeling, and offline reinforcement learning will be introduced and discussed.  

 

Bio

- Education
    - B.S. EE, KAIST, Aug, 2004.
    - Ph.D. EE, KAIST, Aug, 2011. (Supervisor: Chang D. Yoo, Research Area: Machine Learning)
    - Research Internship, National ICT Australia (NICTA), w/ Alex Smola, July 2008.
    - Research Internship, Microsoft Research Cambridge, w/ Pushmeet Kohli and Sebastian Nowozin, May 2010.
- Employment
    - Post Doc., KAIST, Sep. 2011 – Mar. 2012.
    - Staff Engineer, Qualcomm Research Korea, Mar. 2012 – Aug. 2017.
    - Research Scientist, Kakao Brain, Sep. 2017 – Present.
I have got BS and Ph.D degrees from KAIST in 2004 and 2011, respectively. When I was a graduate student, my research area was machine learning, especially applied to speech and image processing, under supervision by Professor Chang D. Yoo. In the middle of my Ph.D course, I performed research internships at National ICT Australia under supervision by Dr. Alex Smola and Mircosoft Research Cambridge under supervision by Dr. Pushmeet Kohli and Dr. Sebastian Nowozin. After receiving my Ph.D degree, I was in KAIST as a postdoc researcher for 6 months, and then worked as a staff engineer at Qualcomm Research Korea for five and half years. In 2017, I joined Kakao Brain and now I am working as a research scientist conducting several research projects mostly related to artificial general intelligence.
 
 
 
최종현 교수 (연세대학교)
 
Title: Continual Learning in Practical Scenarios
 

Abs

Continual learning, especially class-incremental learning uses an episodic memory for past knowledge for better performance. Updating a model with the episodic memory is similar to (1) updating a model with past knowledge in the memory by a few-shot learning scheme, and (2) learning an imbalanced distribution of past data and the present data. We address the unrealistic factors in popular continual learning setups and propose a few ideas to make the continual learning research in realistic scenarios. 

 

Bio

Jonghyun received the B.S. and M.S. degrees in electrical engineering and computer science from Seoul National University, Seoul, South Korea in 2003 and 2008, respectively. He received a Ph.D. degree from University of Maryland, College Park in 2015, under the supervision of Prof. Larry S. Davis. He is currently an associate professor at Yonsei University, Seoul, South Korea. During his PhD, he has worked as a research intern in US Army Research Lab (2012), Adobe Research (2013), Disney Research Pittsburgh (2014) and Microsoft Research Redmond (2014). He was a senior researcher at Comcast Applied AI Research, Washington, DC from 2015 to 2016. He was a research scientist at Allen Institute for Artificial Intelligence (AI2), Seattle, WA from 2016 to 2018 and is currently an affiliated research scientist. He was an assistant professor at GIST, South Korea. His research interest includes visual recognition using weakly supervised data for semantic understanding of images and videos and visual understanding for edge devices and household robots.
 
 
정준원 교수 (가천대학교 길병원)         김광기 교수 (가천대학교)
 
Title:  Application artificial intelligence for gastrointestinal endoscopy
 

Abs

Stomach cancer is the third leading cause of global cancer mortality. Early detection treatment remains the best measure to improve patient survival. Early gastric cancer (EGC) is hard to find and so it can be easily overlooked. Currently, screening for EGC is based on direct visualization during gastroscopy. Meticulous examination of the whole stomach using current techniques can be time-consuming. Since the fact that early cancer detection significantly improves the prognosis, the need for reliable detection-systems of EGC is increasing recently. In this tutorial, how the application of AI to endoscopy could help endoscopist to detect cancer and improve survival or quality of life for patients

 

Bio (정준원 교수):

Graduated Kyung Hee University. M.D.
Ulsan University Ph.D
Currently Head of gastroenterology, Gachon University, Gil Medical Center
CEO of CAIMI
 
Bio (김광기 교수):
- 학력
포항공과대학교 이과대학 물리학과 이학석사
서울대학교 협동과정 의용생체공학 공학박사
- 경력
2007.02-현재  가천대학교 의과대학 의예과 부교수
2007.07-2017.01  국립암센터 연구소 의공학연구과 책임연구원 
1999-2007.06  서울대학교 의학연구원 전임대우 연구조교수 
 
 
 
 
정준선 교수 (KAIST)
 
Title: Self-supervised learning of audio and speech representations
 

Abs

Supervised learning with deep neural networks has brought phenomenal advances to many fields of research, but the performance of such systems relies heavily on the quality and quantity of annotated databases tailored to the particular application. It can be prohibitively difficult to manually collect and annotate databases for every task. There is a plethora of data on the internet that is not used in machine learning due to the lack of such annotations. Self-supervised learning allows a model to learn representations using properties inherent in the data itself, such as natural co-occurrence.

In this talk, I will introduce recent works on self-supervised learning of audio and speech representations. Recent works demonstrates that phonetic and semantic representations of audio and speech can be learnt from unlabelled audio and video. The learnt representations can be used for downstream tasks such as automatic speech recognition, speaker recognition, face recognition and lip reading. Other noteworthy applications include localizing sound sources in images and separating simultaneous speech from video.  

 

Bio

Joon Son Chung is an assistant professor at the School of Electrical Engineering, KAIST, where he is directing Multimodal AI Lab. Previously, he was a research scientist at Naver Corporation, where he managed the development of speech recognition models for various applications including Clova Note. He received his BA and PhD from the University of Oxford, working with Prof. Andrew Zisserman. He published in top journals including TPAMI and IJCV, and has been the recipient of best paper awards at Interspeech and ACCV. His research interests include speaker recognition, cross-modal learning, visual speech synthesis and audio-visual speech recognition.  
 

 
 
박호건 교수 (성균관대학교)
 
Title: GNN의 기본 개념 및 최신 동향
 

Abs

그래프 데이터는 소셜 네트워크, 통신, 단백질 상호작용 등 자연과학 및 사회과학 전반에 걸쳐 폭넗게 활용되고 있다. Graph Neural Networks(GNNs)은 그래프 데이터를 인풋으로 사용하는 그래프 표현 학습(Graph Representation Learning) 방법 중에 하나로, 노드 분류, 링크 분석, 클러스터링 등 다양한 그래프 활용 응용 전반에서 기존 방법론 대비 높은 성능을 보여주고 있다. 본 튜토리얼에서는 GNN의 기본 개념, 이론적 동기, 실제 적용 사례들을 공유하고, 몇 가지 주목할만한 최신 동향을 소개하고자 한다. 

 

Bio

2020-현재 성균관대학교 소프트웨어학과 조교수
2020 미국 Purdue University 컴퓨터과학학과 박사
2009-2013 KIST 영상미디어연구센터 연구원