BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:복소 산술 신경망을 통한 물리적 기계학습모형의 
 건설과 물리적 불변량의 추출
DTSTART;VALUE=DATE-TIME:20191108T074000Z
DTEND;VALUE=DATE-TIME:20191108T075500Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-3@sshep.snu.ac.kr
DESCRIPTION:Speakers: Won Sang Cho (Seoul National University)\nhttps://ss
 hep.snu.ac.kr/event/107/contributions/3/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/3/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Hoping for Better AI: Advice from the Brain
DTSTART;VALUE=DATE-TIME:20191108T002000Z
DTEND;VALUE=DATE-TIME:20191108T004000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-4@sshep.snu.ac.kr
DESCRIPTION:Speakers: Inah Lee (Seoul National University)\nArtificial int
 elligence (AI)\, by definition\, is an artificial version of natural intel
 ligence (NI). Trying to understand how natural intelligence works\, Alan T
 uring\, Demis Hassabis and other AI experts have tried to understand how t
 he human brain works. With the recent explosive advance in technology\, mo
 dern neuroscience is now getting closer to unraveling the mysterious algor
 ithms of the brain one by one. Some of the algorithmic principles as to ho
 w the brain processes information (e.g.\, reinforcement learning\, deep ne
 ural networks with backpropagation\, etc.) have been readily adopted by th
 e AI research field and have revolutionized the field as evidenced by the 
 famous ‘AlphaGo’ by the Google’s DeepMind. However\, the current ver
 sion of AI is inferior to the brain in many aspects although neuroscience 
 has a lot more to offer in pushing the envelope in AI technology to make A
 I virtually inseparable from NI in the future. In this short talk\, I will
  introduce some of the key areas in which modern AI to improve to resemble
  the human brain.\n\nhttps://sshep.snu.ac.kr/event/107/contributions/4/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/4/
END:VEVENT
BEGIN:VEVENT
SUMMARY:What Can We Learn with ML in Particle Physics\, and How?
DTSTART;VALUE=DATE-TIME:20191107T064500Z
DTEND;VALUE=DATE-TIME:20191107T070500Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-5@sshep.snu.ac.kr
DESCRIPTION:Speakers: Sunghoon Jung (Seoul National University)\nML has a 
 deep potential to extend our ability to understand Nature\, beyond common 
 knowledge. As the very first steps to realize it\, we use ML to seek for a
 nswers to one of the not-well-solved problems in particle physics. Our foc
 us is not only to improve the solution\, but to develop ways to figure out
  what the network has learned.\n\nhttps://sshep.snu.ac.kr/event/107/contri
 butions/5/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/5/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Searching for Axions and new concepts in fundamental physics with 
 Machine Learning
DTSTART;VALUE=DATE-TIME:20191107T070500Z
DTEND;VALUE=DATE-TIME:20191107T073500Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-6@sshep.snu.ac.kr
DESCRIPTION:Speakers: Sven Krippendorf (Ludwig Maximilian University of Mu
 nich)\nIn this talk\, I will review how ultralight axion-like particles ca
 n be constrained using X-ray observations of bright localized sources (AGN
 s\, Quasars) in and behind galaxy clusters. To find axion-like particles i
 n these settings corresponds to finding their characteristic pattern (quas
 i-sinusoidal oscillations) in noisy data. Using ML-techniques we are able 
 to improve the search sensitivity for these particles and are able to set 
 stronger bounds compared to previous methods. In the second part of the ta
 lk\, I shall highlight avenues where ML can be used to accelerate our unde
 rstanding of fundamental physics (string theory) and potentially vice vers
 a.\n\nhttps://sshep.snu.ac.kr/event/107/contributions/6/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/6/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Welcome remarks on the 1st XAIENCE Conference
DTSTART;VALUE=DATE-TIME:20191107T000000Z
DTEND;VALUE=DATE-TIME:20191107T002000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-7@sshep.snu.ac.kr
DESCRIPTION:Speakers: Junho Lee (Seoul National University)\nhttps://sshep
 .snu.ac.kr/event/107/contributions/7/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/7/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Application of Machine Learning for Big Data Analysis in Astronomy
  (천문학 자료분석에서의 기계학습의 활용)
DTSTART;VALUE=DATE-TIME:20191107T055000Z
DTEND;VALUE=DATE-TIME:20191107T061000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-8@sshep.snu.ac.kr
DESCRIPTION:Speakers: Min-Su Shin (KASI)\n천문학 관측 자료의 분
 석의 경우\, 관측 환경의 변화 등에 의해서 야기된 관측 
 부정확성이나 편향성\, 그리고 비균질적인 자료 획득 등
 의 특징적인 문제를 가지고 있다. 이러한 천문 관측 자
 료의 특성을 고려해서\, 천문학자들은 천체나 천체 현상
 의 검출\, 분류\, 특성의 추론 등의 문제를 접하게 된다. 
 이러한 문제들은  관측 자료의 급속한 증가와 함께 더욱
  심각해지는데\, 기계학습의 활용을 통하여 정량적인 검
 출\, 분류\, 추론의 문제들을 해결해 가고자 노력 중이다
 . 이번 발표에서는 이러한 문제들을 clustering\, anomaly detec
 tion\, classification\, regression\, recommendation 문제의 시각에서
  기계학습을 활용한 사례들을 소개하고\, 최근 시도되고
  있는 ensemble learning\, multi-task learning 등의 활용도 간랸히
  제시하고자 한다. 더불어서 천체와 그 현상에 대한 천
 체물리적 이해를 위한 기계학습의 활용 방향에 대해서 
 의견을 제시한다.\n\nhttps://sshep.snu.ac.kr/event/107/contributions
 /8/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/8/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Machine Learning In Astrophysics:  Estimating Galactic Baryonic Pr
 operties from Their Dark Matter
DTSTART;VALUE=DATE-TIME:20191107T061000Z
DTEND;VALUE=DATE-TIME:20191107T063000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-9@sshep.snu.ac.kr
DESCRIPTION:Speakers: Ji-hoon Kim (SNU)\nAs astrophysics and cosmology dea
 l with inherently "cosmological" sizes of data\, machine learning is being
  rapidly adopted in a variety of astrophysical applications.  In this talk
 \, I will introduce a pipeline that estimates baryonic (visible) propertie
 s of a galaxy based purely on dark matter (DM\; invisible) properties in l
 arge-scale DM-only simulations.  It is shown that our pipeline promptly ge
 nerates a galaxy catalogue from a DM halo catalogue using a machine traine
 d on smaller-scale\, fully-hydrodynamic\, high-resolution simulations. An 
 extremely randomized tree algorithm is used together with multiple novel i
 mprovements we developed such as a refined error function and two-stage le
 arning.  Our model may become a promising method to transplant the baryon 
 physics of galaxy-scale hydrodynamic calculations onto a larger-volume DM-
 only run.\n\nhttps://sshep.snu.ac.kr/event/107/contributions/9/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/9/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Recent Developments of Deep Learning in Experimental High Energy P
 hysics
DTSTART;VALUE=DATE-TIME:20191107T075500Z
DTEND;VALUE=DATE-TIME:20191107T081500Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-10@sshep.snu.ac.kr
DESCRIPTION:Speakers: Hwidong Yoo (Seoul National University)\n최근들
 어 딥러닝 기술은 기초 과학 분야에도 광범위하게 연구
 되고 있다. 이를 통해 얻어진 지식 및 기술이 차세대 연
 구 시스템 및 환경을 변화시키고 결과를 획기적으로 향
 상시키기 위해 다양하게 적용될 것으로 기대되고 있다. 
 현재 고에너지 물리 실험분야에서 활발하게 연구되고 
 있는 딥러닝 관련 연구에 대해 소개하고 향후 기대 효과
 에 대해 논의 한다.\n\nhttps://sshep.snu.ac.kr/event/107/contributio
 ns/10/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/10/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Learning QCD Jet Flavors in search for Invisible Higgs Decays
DTSTART;VALUE=DATE-TIME:20191107T073500Z
DTEND;VALUE=DATE-TIME:20191107T075500Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-11@sshep.snu.ac.kr
DESCRIPTION:Speakers: Hyung Do Kim (Seoul National University)\nhttps://ss
 hep.snu.ac.kr/event/107/contributions/11/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/11/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Connectome Project & Neuroscience with AI
DTSTART;VALUE=DATE-TIME:20191108T000000Z
DTEND;VALUE=DATE-TIME:20191108T002000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-12@sshep.snu.ac.kr
DESCRIPTION:Speakers: Junho Lee (Seoul National University)\n예쁜꼬마
 선충 C. elegans는 신경계의 발생과 진화를 연구할 수 있는
  최적의 모델동물이다. 꼬마선충의 신경계는 302개의 뉴
 런으로 구성된 최소화된 뇌의 모델이라 할 수 있다.\n\n
 본 연구실에서는 nictation 이라고 하는 히치하이킹 행동
 의 신경회로의 작동 및 조절\, 그리고 진화에 대한 연구
 를 수행하고 있다. Nictation 행동은 특정 발생 단계 (dauer
 라는 휴면 단계)의 선충에서만 나타나는 행동으로서 고
 개를 들어 몸을 세우고 필요한 경우 흔드는 행동을 의미
 한다. 본 연구실의 선행연구로 많은 뉴런 중 IL2라는 뉴
 런이 핵심적인 역할을 하는 것을 밝혔고 지구상에 흩어
 져 있는 다른 꼬마선충들이 그 nictation의 정도를 달리하
 는 유전적 다양성을 가짐도 보인 바 있다.\n\n현재 가장 
 중요한 문제는 nictation 행동이 어떻게 dauer라고 하는 휴
 면 유충에서만 실행되는가이고\, 그 답이 될 수 있는 것
 은 신경회로 즉 커넥톰의 차이에 있을 수 있다는 것이다
 . 이 가설을 검증하고 신경계의 발생유연성의 기전을 연
 구하기 위해 우리는 꼬마선충 dauer를 50 나노미터 두께로
  연속절편하여 전자현미경 사진을 찍은 후 모든 3차 구
 조를 재구성하는 작업을 진행하고 있다. 여기서 AI가 관
 여하는 작업이 가능하다면 이룰 수 있는 일들이 획기적
 으로 많아질 것으로 기대하며 본 연구의 진척과 문제점 
 등을 공유하고자 한다.\n\nhttps://sshep.snu.ac.kr/event/107/contri
 butions/12/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/12/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Adjourn
DTSTART;VALUE=DATE-TIME:20191108T075500Z
DTEND;VALUE=DATE-TIME:20191108T081000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-13@sshep.snu.ac.kr
DESCRIPTION:https://sshep.snu.ac.kr/event/107/contributions/13/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/13/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Application of Artificial Intelligence techniques to Weather Forec
 asting
DTSTART;VALUE=DATE-TIME:20191108T020000Z
DTEND;VALUE=DATE-TIME:20191108T022000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-14@sshep.snu.ac.kr
DESCRIPTION:Speakers: Jung-Hoon Kim (Seoul National University)\nhttps://s
 shep.snu.ac.kr/event/107/contributions/14/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/14/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Probing Trilinear Higgs Self–coupling at the HL-LHC with Machine
  Learning
DTSTART;VALUE=DATE-TIME:20191108T064000Z
DTEND;VALUE=DATE-TIME:20191108T065500Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-15@sshep.snu.ac.kr
DESCRIPTION:Speakers: Jubin Park (Institute for Universe and Elementary Pa
 rticles\, Chonnam National University)\n1964년\, 피터 힉스에 의해 
 제안된 힉스 메커니즘을 통해 힉스입자의 존재가 예측
 되었다. 이후 48년이 지나\, 2012년 7월 유럽입자물리연구
 소(CERN) 대형강입자 충돌기(LHC)에서 드디어 이 입자가 발
 견되었다. 현재 이 입자의 성질을 정밀하게 측정하고 검
 증하고자\, LHC에서 다양한 연구들이 진행 중에 있다. 이 
 발표에서는 이 중\, 힉스입자의 삼중 커플링의 검측 성
 능을 머신러닝의 기술을 이용해 개선한 예를 통해\, 고
 에너지 물리학에서의 응용을 소개하고자 한다.\n\nhttps://
 sshep.snu.ac.kr/event/107/contributions/15/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/15/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Delfos: deep learning model for prediction of solvation free energ
 ies in generic organic solvents
DTSTART;VALUE=DATE-TIME:20191108T040000Z
DTEND;VALUE=DATE-TIME:20191108T042000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-16@sshep.snu.ac.kr
DESCRIPTION:Speakers: YounJoun Jung (Seoul National University)\nPredictio
 n of aqueous solubilities or hydration free energies is an extensively stu
 died area in machine learning applications in chemistry since water is the
  sole solvent in the living system. However\, for non-aqueous solutions\, 
 few machine learning studies have been undertaken so far despite the fact 
 that the solvation mechanism plays an important role in various chemical r
 eactions. Here\, we introduce Delfos (deep learning model for solvation fr
 ee energies in generic organic solvents)\, which is a novel\, machine-lear
 ning-based QSPR method which predicts solvation free energies for various 
 organic solute and solvent systems. A novelty of Delfos involves two separ
 ate solvent and solute encoder networks that can quantify structural featu
 res of given compounds via word embedding and recurrent layers\, augmented
  with the attention mechanism which extracts important substructures from 
 outputs of recurrent neural networks. As a result\, the predictor network 
 calculates the solvation free energy of a given solvent–solute pair usin
 g features from encoders. With the results obtained from extensive calcula
 tions using 2495 solute–solvent pairs\, we demonstrate that Delfos not o
 nly has great potential in showing accuracy comparable to that of the stat
 e-of-the-art computational chemistry methods\, but also offers information
  about which substructures play a dominant role in the solvation process.\
 n\nReferences:\n\n1. [Delfos: deep learning model for prediction of solvat
 ion free energies in generic organic solvents\, Hyuntae Lim and YounJoon J
 ung\, Chemical Science 2019 DOI: 10.1039/C9SC02452B (2019)] [1]\n\n2. [Med
 ia] [2]\n\n\n  [1]: https://pubs.rsc.org/en/content/articlelanding/2019/sc
 /c9sc02452b#!divAbstract\n  [2]: http://now.snu.ac.kr/47/3/1450\n\nhttps:/
 /sshep.snu.ac.kr/event/107/contributions/16/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/16/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Application of Neural Network Model to Predict and to Evaluate the
  Groundwater Level Fluctuation
DTSTART;VALUE=DATE-TIME:20191108T072500Z
DTEND;VALUE=DATE-TIME:20191108T074000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-17@sshep.snu.ac.kr
DESCRIPTION:Speakers: Sanghoon Lee (Seoul National University)\nGroundwate
 r is one of valuable water resources used for the various purposes in our 
 life\, but it is not limitless. For this reason\, prediction and managemen
 t of the groundwater are very essential work for sustainable use of it. Ph
 ysics-based models are usually applied on prediction of groundwater level\
 , but they are hard to be implemented successfully when there is any unkno
 wn physical property or when subterranean structure is very complicated. I
 n this research\, prediction of the groundwater level at riverside area in
  Yangpyeong\, Korea was carried out using neural network model instead. In
  study area where several natural and anthropogenic factors affect the gro
 undwater level fluctuation\, groundwater levels at 8 monitoring wells were
  well predicted with low range of RMSE errors. Moreover\, monthly contribu
 tions\, which indicate the impact of input variables\, were computed to fi
 gure out the seasonal variance of influencing factors. This study could su
 ggest another option to predict the groundwater level\, and help understan
 ding spatial and temporal variation of impacts of factors affecting the gr
 oundwater level fluctuation.\n\nAcknowledgement: This work was supported b
 y the National Research Foundation of Korea(NRF) grant funded by the Korea
  government(MSIP) (No. 2017R1A2B3002119)\n\nhttps://sshep.snu.ac.kr/event/
 107/contributions/17/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/17/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Nonconvex Sparse Regularization For Deep Neural Networks and its O
 ptimal Property
DTSTART;VALUE=DATE-TIME:20191107T083000Z
DTEND;VALUE=DATE-TIME:20191107T084500Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-18@sshep.snu.ac.kr
DESCRIPTION:Speakers: Ilsang Ohn (Department of Statistics\, Seoul Nationa
 l University)\nSparsity is a key ingredient in the success of learners bot
 h theoretically and computationally. This is also the case for deep neural
  networks (DNNs). A number of empirical observations show that sparse DNNs
  can dramatically reduce computation time and memory without appreciably h
 arming prediction power. Furthermore\, recent theoretical studies proved t
 hat DNN estimators with a certain sparsity constraint can attain optimal c
 onvergence rates for regression and classification problems. However\, the
 y only considered the empirical risk minimizer under the sparsity constrai
 nt\, where optimization is almost impossible in practice due to its discre
 te nature. In this research\, we propose a novel penalized empirical risk 
 minimization method for estimating sparse DNNs with a scalable computation
  algorithm.  The proposed method yields sparse DNNs that can achieve optim
 al convergence rates of excess risks for various learning problems includi
 ng regression and binary classification. We demonstrate the empirical perf
 ormance of the proposed method and compare it with other competitors for v
 arious benchmark datasets.\n\nhttps://sshep.snu.ac.kr/event/107/contributi
 ons/18/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/18/
END:VEVENT
BEGIN:VEVENT
SUMMARY:On Casting Importance Weighted Autoencoder to an EM Algorithm to L
 earn Deep Generative Models
DTSTART;VALUE=DATE-TIME:20191107T081500Z
DTEND;VALUE=DATE-TIME:20191107T083000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-19@sshep.snu.ac.kr
DESCRIPTION:Speakers: Dongha Kim (Seoul National University)\nWe propose a
  new and general approach to learn deep generative models. \n  Our approac
 h is based on a new observation that the importance weighted autoencoders 
 (IWAE) can be understood as a procedure of estimating the MLE with an EM a
 lgorithm. \n  Utilizing this interpretation\, we develop a new learning al
 gorithm called importance weighted EM algorithm (IWEM). \nIWEM is an EM al
 gorithm with importance sampling (IS)\nwhere the proposal distribution is 
 carefully selected to reduce the variance\ndue to IS. In addition\, we dev
 ise an annealing strategy to stabilize the learning algorithm. For missing
  data problems\, we propose a modified\nIWEM algorithm called miss-IWEM. \
 n  Using multiple benchmark datasets\, we demonstrate empirically that our
  proposed methods outperform IWAE with significant margins for both fully-
 observed and missing data cases.\n\nhttps://sshep.snu.ac.kr/event/107/cont
 ributions/19/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/19/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Theoretical Advantages of Deep Neural Networks
DTSTART;VALUE=DATE-TIME:20191107T002000Z
DTEND;VALUE=DATE-TIME:20191107T004000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-20@sshep.snu.ac.kr
DESCRIPTION:Speakers: Yongdai Kim (Seoul National University)\nFrom statis
 tical points of view\, deep neural networks (DNN) are nothing but a (gener
 alized) regression model\, but DNNs have solved many problems no other met
 hods have not succeeded in the past. In this talk\, I will explain theoret
 ical advantages of DNNs compared to other nonparametric regression models.
 \n\nhttps://sshep.snu.ac.kr/event/107/contributions/20/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/20/
END:VEVENT
BEGIN:VEVENT
SUMMARY:AI Applications for Satellite-based Disaster Monitoring and Predic
 tion
DTSTART;VALUE=DATE-TIME:20191107T023000Z
DTEND;VALUE=DATE-TIME:20191107T025000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-29@sshep.snu.ac.kr
DESCRIPTION:Speakers: Jungho Im (UNIST)\nhttps://sshep.snu.ac.kr/event/107
 /contributions/29/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/29/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Deep learning for multi-year ENSO forecasts
DTSTART;VALUE=DATE-TIME:20191107T021000Z
DTEND;VALUE=DATE-TIME:20191107T023000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-21@sshep.snu.ac.kr
DESCRIPTION:Speakers: Yoo-Geun Ham (Chonnam National University)\nVariatio
 ns in the El Niño/Southern Oscillation (ENSO) are associated with a wide 
 array of regional climate extremes and ecosystem impacts. Robust\, long-le
 ad forecasts would therefore be valuable for managing policy responses. Bu
 t despite decades of effort\, forecasting ENSO events at lead times of mor
 e than one year remains problematic. Here we show that a statistical forec
 ast model employing a deep-learning approach produces skilful ENSO forecas
 ts for lead times of up to one and a half years. To circumvent the limited
  amount of observation data\, we use transfer learning to train a convolut
 ional neural network (CNN) first on historical simulations and subsequentl
 y on reanalysis from 1871 to 1973. During the validation period from 1984 
 to 2017\, the all-season correlation skill of the Nino3.4 index of the CNN
  model is much higher than those of current state-of-the-art dynamical for
 ecast systems. The CNN model is also better at predicting the detailed zon
 al distribution of sea surface temperatures\, overcoming a weakness of dyn
 amical forecast models. A heat map analysis indicates that the CNN model p
 redicts ENSO events using physically reasonable precursors. The CNN model 
 is thus a powerful tool for both the prediction of ENSO events and for the
  analysis of their associated complex mechanisms.\n\n\nReferences:  \n\n[h
 ttps://www.nature.com/articles/s41586-019-1559-7][1]\n\n[https://www.scien
 cemag.org/news/2019/09/artificial-intelligence-could-predict-el-ni-o-18-mo
 nths-advance][2]\n\n[https://www.yna.co.kr/view/AKR20190918093900017?input
 =1195m][3]\n\n[http://biz.chosun.com/site/data/html_dir/2019/09/19/2019091
 900659.html][4]\n\n\n  [1]: https://www.nature.com/articles/s41586-019-155
 9-7\n  [2]: https://www.sciencemag.org/news/2019/09/artificial-intelligenc
 e-could-predict-el-ni-o-18-months-advance\n  [3]: https://www.yna.co.kr/vi
 ew/AKR20190918093900017?input=1195m\n  [4]: http://biz.chosun.com/site/dat
 a/html_dir/2019/09/19/2019091900659.html\n\nhttps://sshep.snu.ac.kr/event/
 107/contributions/21/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/21/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Prediction and Taking Insight of Molecular Quantum Property by Ran
 dom Forest Regression
DTSTART;VALUE=DATE-TIME:20191108T065500Z
DTEND;VALUE=DATE-TIME:20191108T071000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-22@sshep.snu.ac.kr
DESCRIPTION:Speakers: Beomchang Kang (Seoul National University)\nFluoresc
 ent molecules are widely used for bio-imaging. They are attached to specif
 ic cell organelles or proteins\, enabling observation of detailed structur
 e and dynamics in the cell. Efficient fluorescent molecules must have a hi
 gh quantum yield for effective bio-imaging. Diverse effective fluorescent 
 molecules whose color are distinctive need to get more information of cell
  or protein. First step of discover novel molecules using computational ap
 proach is prediction of compound. Here\, we use random forest regression t
 o predict excitation energy and oscillator strength of a molecule. Random 
 forest algorithm is white box. It is easy to extract feature importance. W
 e could get insight from it.\nA statistical machine that predicts excitati
 on energies and associated oscillator strengths\, the probability of absor
 ption or emission of light in transitions between different energy states\
 , of a molecule were trained using the random forest algorithm. The Pubche
 mQC database was used as a training set. It has over 3 million known compo
 unds. We picked up 0.5 million molecules  randomly. 90% of them were in tr
 aining set and the others were test set. The ECFP4(extended connectivity f
 ingerprints 2) of molecules were used as input features. We found some fra
 gments which can decide molecules’ quantum property by feature importanc
 e analysis.\n\nhttps://sshep.snu.ac.kr/event/107/contributions/22/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/22/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Predicting location of suitable groundwater for brewing coffee usi
 ng tree-based ensemble machine learning
DTSTART;VALUE=DATE-TIME:20191108T071000Z
DTEND;VALUE=DATE-TIME:20191108T072500Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-23@sshep.snu.ac.kr
DESCRIPTION:Speakers: Hye-Lim Lee (School of Earth and Environmental Scien
 ces\, Seoul National University)\nRecently\, groundwater is used as a sour
 ce of drinking purposes such as water\, coffee\, beer and other beverages.
  The range of water quality for producing high-quality of beverages is dif
 ferent from each usage\, so it is important to find suitable groundwater l
 ocation for each purpose in the aspect of water industry. This study was c
 onducted to predict the suitable location for brewing coffee in Gangwon Pr
 ovince\, South Korea using tree-based ensemble machine learning. Appropria
 te water quality standard for brewing coffee is known as TDS of 75~250 mg/
 L and calcium hardness of 17~85 mg/L from recent research. Boosted Regress
 ion Trees (BRT)\, Random Forests (RF) and Extremely Randomized Trees (ERT)
  were used as tree-based ensemble method. Response indicating suitable or 
 unsuitable groundwater for brewing coffee was determined by 254 wells’ w
 ater quality data\, and predictor variables were composed of slope\, altit
 ude\, drainage grade\, effective soil depth\, soil composition\, land use\
 , and hydrogeology based on GIS data. Applying models to the test data\, a
 ll three models showed the area under a curve (AUC) and accuracy more than
  0.85 and 0.80\, respectively\, which indicates high reliability in predic
 tion. Threshold of dividing suitable or unsuitable for brewing coffee is f
 ound from Receiver Operating Characteristic (ROC) curve. The BRT showed th
 e highest AUC and accuracy among the three models\, therefore\, potential 
 map of suitable groundwater location for brewing coffee was suggested by t
 he BRT model. In the condition of lack of water quality data\, this resear
 ch can help to determine location of suitable groundwater for several usag
 es.\nKeyword: potable groundwater · boosted regression tree · random for
 est · extremely randomized tree \nAcknowledgement: This research was supp
 orted by the National Research Council of Science and Technology(NST) gran
 t funded by the Korea government(MSIP) (No. CAP-17-05-KIGAM)\n\nhttps://ss
 hep.snu.ac.kr/event/107/contributions/23/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/23/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Clustering weather maps for high PM10 events in Seoul using Self O
 rganization Maps
DTSTART;VALUE=DATE-TIME:20191108T022000Z
DTEND;VALUE=DATE-TIME:20191108T024000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-24@sshep.snu.ac.kr
DESCRIPTION:Speakers: Seok-Woo Son (Seoul National University)\nhttps://ss
 hep.snu.ac.kr/event/107/contributions/24/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/24/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Projection onto Minkowski Sums with Application to Constrained Lea
 rning
DTSTART;VALUE=DATE-TIME:20191107T004000Z
DTEND;VALUE=DATE-TIME:20191107T010000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-25@sshep.snu.ac.kr
DESCRIPTION:Speakers: Joong-Ho Won (Seoul National University)\nWe introdu
 ce block descent algorithms for projecting onto Minkowski sums of sets. Pr
 ojection onto such sets is a crucial step in many statistical learning pro
 blems\, and may regularize complexity of solutions to an optimization prob
 lem or arise in dual formulations of penalty methods. We show that project
 ing onto the Minkowski sum admits simple\, efficient algorithms when compl
 ications such as overlapping constraints pose challenges to existing metho
 ds. We prove that our algorithm converges linearly when sets are strongly 
 convex or satisfy an error bound condition\, and extend the theory and met
 hods to encompass non-convex sets as well. We demonstrate empirical advant
 ages in runtime and accuracy over competitors in applications to ℓ1\,p-r
 egularized learning\, constrained lasso\, and overlapping group lasso.\n\n
 \nReference:\n1. Joong-Ho Won\, Jason Xu\, Kenneth Lange\; [Proceedings of
  the 36th International Conference on Machine Learning\, PMLR 97:3642-3651
 \, 2019][1]. \n\n\n  [1]: http://proceedings.mlr.press/v97/lange19a.html\n
 \nhttps://sshep.snu.ac.kr/event/107/contributions/25/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/25/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Deep Learning for Sea Ice Classification and Ship Detection using 
 Satellite Synthetic Aperture Radar
DTSTART;VALUE=DATE-TIME:20191107T015000Z
DTEND;VALUE=DATE-TIME:20191107T021000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-26@sshep.snu.ac.kr
DESCRIPTION:Speakers: Duk-jin Kim (Seoul National University)\nhttps://ssh
 ep.snu.ac.kr/event/107/contributions/26/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/26/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Lessons Learned from Running Interdisciplinary Computational Scien
 ces Program for Undergraduate at Seoul National University on the Future o
 f Scientific Computing Education
DTSTART;VALUE=DATE-TIME:20191107T012000Z
DTEND;VALUE=DATE-TIME:20191107T015000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-27@sshep.snu.ac.kr
DESCRIPTION:Speakers: Sang-Mook Lee (Seoul National University)\n자연과
 학은 자연을 대상으로 한 관측이 주된 목표이고 우리 자
 연과학자들은 관측자료 즉 데이터를 통해 자연(모델)을 
 이해하는 것에 매우 익숙하다. 그런데 최근 갑자기 데이
 터과학\, 인공지능이며 머신러닝이 4차 산업혁명이라는 
 buzzword와 함께 큰 화제이다. 기업(enterprise) 차원에서는 
 이것들이 새로운 분야이자 기회인 것 처럼 보일지 모르
 지만 아직 자연과학 분야에서의 기여도는 제한적이다. 
 정말 인공지능 머신러닝을 통해 새로운 자연현상이 밝
 혀진 사례는 최소한 내가 속한 지질해양분야에서는 그
 리 많은 것 같지 않다. 그럼에도 불구하고 학부생들과 
 대학원생들에게 새로운 분야를 가르쳐야하는 교육자로
 써 지난 5-6 년간 교육현장에서 느낀 점들을 공유하고자 
 한다. 특히 enterprise AI와 달리 scientific AI가 나아가야 할 
 방향 그리고 컴퓨터과학의 발달로 인해 달라진 ICT의 생
 태계 속에서 어떻게 미래문제해결형 학생을 양성하기 
 위한 교육이 이루어져야 하는가에 대해 의견을 제시하
 고자 한다. 이는 현재 서울대 계산과학 연합전공과 협동
 과정이 추구하는 방향이기도 하다. \n\nScience is all about ob
 servation and measurement. So for natural scientists like us we should be 
 very comfortable about how to derive our understanding into models from ob
 served data sets. This should be very familiar to us and It is the crux of
  what we do day to day. Recently\, however\, we are puzzled by the new buz
 zword ‘The Fourth Industrial Revolution’ as well as terms such as data
  science\, machine learning and artificial intelligence. However\, at leas
 t in my field of marine and Earth sciences\, the impact brought about by t
 hese new disciplines and approach have been somewhat limited because in sc
 ience we never get enough data as we want. These new approaches require im
 mense amounts of data\, which sometimes is not practical. As a person who 
 has been involved as an educator in the midst of these new changes\, I wou
 ld like to share my views on how we can the important differences between 
 enterprise AI and scientific approaches. It is important to teach students
  probably under these new changes because our goal has always been to rais
 e competent next-generation scientists to deal with future problems. I hop
 e I can suggest a better picture of scientific computation in this world o
 f AI which we had computational sciences program of Seoul National Univers
 ity is also trying to find.\n\nhttps://sshep.snu.ac.kr/event/107/contribut
 ions/27/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/27/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Prediction of Protein Structure and Interaction by Physics and Inf
 ormatics
DTSTART;VALUE=DATE-TIME:20191107T043000Z
DTEND;VALUE=DATE-TIME:20191107T045000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-28@sshep.snu.ac.kr
DESCRIPTION:Speakers: Chaok Seok (Seoul National University)\nProtein stru
 cture prediction problem has challenged theoretical and computational phys
 ical scientists since the first protein structure was published in 1958. T
 here have been steady progresses in protein structure prediction since the
 n\, but major contributions to the progress came from informatics-based ap
 proaches rather than from physics-based approaches. Recently\, DeepMind’
 s AlphaFold made a further contribution by introducing deep learning to ex
 tract structural information from the large sequence database. Meaningful 
 contribution of physics-based approaches began to be made only in 2012 in 
 the field of structure refinement. However\, structural improvements that 
 can be achieved by refinement with current physics-based approaches are ve
 ry limited due to both energy and sampling problems. To overcome this limi
 tation\, we are taking an approach that combines physics and informatics\,
  including deep learning. We take similar approaches to predict interactio
 ns of proteins with other proteins or small ligands including short peptid
 es and oligosaccharides. Our goal is to develop protein structure modeling
  techniques that can provide useful predictions even in the absence of ava
 ilable information\, although currently available experimental data would 
 play important roles in developing such techniques. Such modeling methods 
 would be very useful for applications to a wide range of biomedical resear
 ch and drug discovery.\n\nhttps://sshep.snu.ac.kr/event/107/contributions/
 28/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/28/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Fleming : AI-driven Integrated Drug Discovery Platform
DTSTART;VALUE=DATE-TIME:20191107T051000Z
DTEND;VALUE=DATE-TIME:20191107T053000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-30@sshep.snu.ac.kr
DESCRIPTION:Speakers: Junsu Ko (© Arontier)\nNew drug development cost mo
 re than one billion over the last decade. And the chance of success was qu
 ite low\, one in five thousands. In order to make the drug development pro
 cess efficient in terms of time and cost\, Artificial Intelligence has bee
 n actively introduced to and rigorously adopted in the various fields of d
 rug development: compound activity prediction\, compound design\, patient 
 selection\, and clinical trial design\, just to name a few.\n\nAs these fi
 elds become advanced\, more AI-based drug development platforms are develo
 ped and deployed in the fields to overcome data shortage.\n\nFleming is an
  integrated and automated platform for time and cost efficient drug develo
 pment. It uses protein structure prediction and genome-based target select
 ion for efficiency. In Fleming\, precise target protein structures\, gener
 ated by protein structure prediction techniques using AI\, are used to pic
 k candidate compounds and to design active compounds with high chance of s
 uccess. And genome-based analysis powered by Fleming AI selects (find-an-a
 djective-for-this) compounds by predicting compound activity and toxicity.
 \n\nThese features in Fleming accelerate the new drug development process 
 utilizing and combining various technologies: disease genome analysis\, ca
 ndidate compound selection\, compound generation\, activity prediction\, a
 nd toxicity prediction.\n\nhttps://sshep.snu.ac.kr/event/107/contributions
 /30/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/30/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Discovering Novel Fluorescent Molecules by Combining Machine-learn
 ing and Global Optimization
DTSTART;VALUE=DATE-TIME:20191107T045000Z
DTEND;VALUE=DATE-TIME:20191107T051000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-31@sshep.snu.ac.kr
DESCRIPTION:Speakers: Juyong Lee (Kangwon National University)\nFluorescen
 t molecules are widely used for bio-imaging. They are attached to specific
  cell organelles and/or proteins\, enabling observation of detailed struct
 ure and dynamics in the cell. Efficient fluorescent molecules must have a 
 high quantum yield for effective bio-imaging. Here\, we present a systemat
 ic approach to discovering novel fluorescent molecules that combines machi
 ne-learning and global optimization algorithms. We recast the problem of d
 iscovering novel fluorescent molecules with high-intensity emission light 
 into a global optimization problem by using the oscillator strength of a m
 olecule as an objective function for optimization.\n\nA statistical machin
 e that predicts excitation energies and associated oscillator strengths\, 
 the probability of absorption or emission of light in transitions between 
 different energy states\, of a molecule were trained using the random fore
 st algorithm. The Pub-chemQC database\, which contains TD-DFT calculation 
 results of 3.8 million known molecules\, was used as a training set. The e
 xtended connectivity fingerprints of molecules were used as input vectors.
  To optimize the oscillator strength of a molecule\, a highly efficient gl
 obal optimization algorithm called CSA was used. For CSA global optimizati
 on\, SMILES representation of a molecule was mapped to a 200-dimensional i
 nteger vector by using Natural Language Toolkit. After CSA global optimiza
 tion calculation converged\, we assessed the validity of our approach by p
 erforming quantum mechanical calculations. TD-DFT calculations were carrie
 d out to verify whether novel molecules obtained by this procedure actuall
 y have high oscillator strength.\n\nhttps://sshep.snu.ac.kr/event/107/cont
 ributions/31/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/31/
END:VEVENT
BEGIN:VEVENT
SUMMARY:데이터의 프라이버시를 보존하는 기계학습 (Safe Ma
 chine Learning toward Private AI)
DTSTART;VALUE=DATE-TIME:20191108T010000Z
DTEND;VALUE=DATE-TIME:20191108T012000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-32@sshep.snu.ac.kr
DESCRIPTION:Speakers: Jung Hee Cheon (Seoul National University)\n기계
 학습(Machine Learning)은 데이터로부터 이를 도출한 함수를 
 유추하는 과정으로 최근 다양한 분야에서 흥미로운 응
 용들이 제시되고 있다. 기계학습이 좋은 성과를 거두려
 면 데이터의 확보가 필수적인데 개인 프라이버시 문제 
 혹은 데이터 주권의 문제로 인해 좋은 데이터를 확보하
 는 것이 쉽지 않은 일이다. 동형암호는 암호화한 데이터
 상에서 복호화없이 기계학습의 훈련단계(Training)나 예측
 단계(Inference)를 수행할 수 있도록 하며\, 이를 통해 데이
 터의 프라이버시 문제를 극복하고 Private AI의 시대를 열
 어가고 있다. 본 강연에서는 이 분야의 최근 결과들로 
 동형 회귀분석(Homomorphic Logistric Regression)\, 동형 심층신
 경망(Homomorphic Deep Neural Network)\, 동형 의사결정나무(Homomo
 rphic Decision Tree)등의 결과와 이의 신용정보\, 의료\, 마케
 팅 등에의 응용을 소개하도록 한다.\n\nhttps://sshep.snu.ac.kr
 /event/107/contributions/32/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/32/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Architecture of Neuromorphic Hardware with Embedded Learning
DTSTART;VALUE=DATE-TIME:20191108T050000Z
DTEND;VALUE=DATE-TIME:20191108T052000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-33@sshep.snu.ac.kr
DESCRIPTION:Speakers: Doo Seok Jeong (Hanynag University)\nhttps://sshep.s
 nu.ac.kr/event/107/contributions/33/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/33/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Quantum Algorithms and Quantum Machine Learning
DTSTART;VALUE=DATE-TIME:20191108T060000Z
DTEND;VALUE=DATE-TIME:20191108T062000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-34@sshep.snu.ac.kr
DESCRIPTION:Speakers: Jeongho Bang (KIAS)\n정보처리/컴퓨팅에의 
 초소형화/고집적화 그리고 무어의 법칙 등은 정보이론
 과 양자물리학의 융합 및 발전이 필연임을 반영한다. 양
 자물리학에 기반한 컴퓨팅에의 속도향상 등의 연구는 
 그 가능성을 증명하는 수준을 넘어\, 양자기술의 실현이
 라는 현실적 목표앞에 와 있다. 이제\, 양자컴퓨팅/통신 
 등의 용어는 학계를 넘어 이미 우리 일상에 와 있다. 이
 에\, 본 강연에서는 기본적인 양자원리\, 즉 양자중첩 및
  얽힘 등을 이해하고\, 알려진 유명한 양자알고리즘들과
  더불어 어떻게 양자머신러닝에의 속도향상이 가능한지
  간략히 살펴보고자 한다. 아울러\, 최근 연구 동향 및 
 이슈 등 또한 간략히 소개하고자 한다.\n\nhttps://sshep.snu.a
 c.kr/event/107/contributions/34/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/34/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Exploration of energy surfaces and conformation of molecules and s
 olids by utilizing a machine-learning technique
DTSTART;VALUE=DATE-TIME:20191108T054000Z
DTEND;VALUE=DATE-TIME:20191108T060000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-35@sshep.snu.ac.kr
DESCRIPTION:Speakers: Jaejun Yu (Seoul National University)\nPredicting th
 e physical properties of novel materials requires an accurate description 
 of atomic interactions as provided by first-principles quantum mechanical 
 calculations. Efficient and practical calculation tools have been develope
 d along with the progress of density functional theory (DFT). Still\, howe
 ver\, the computational complexity associated with the quantum mechanical 
 treatment limits their applications to systems of a few hundreds of atoms 
 at most. Here\, we present an application of the Gaussian process regressi
 on (GPR) scheme to the global optimization\, conformation space annealing\
 , and pathway optimization methods. We demonstrate that the use of GPR-bas
 ed pathway optimization technique\, e.g.\, action-derived molecular dynami
 cs (ADMD) method\, can be useful in enhancing the computational performanc
 e. We will discuss possible future applications of the GPR-based machine l
 earning technique for the exploration of energy surfaces and conformation 
 of molecules and solids.\n\nhttps://sshep.snu.ac.kr/event/107/contribution
 s/35/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/35/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Information in Unsupervised Learning
DTSTART;VALUE=DATE-TIME:20191108T052000Z
DTEND;VALUE=DATE-TIME:20191108T054000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-36@sshep.snu.ac.kr
DESCRIPTION:Speakers: Junghyo Jo (Seoul National University)\nInformation 
 bottleneck theory explains that neural networks maximally compress unneces
 sary information in input data\, and transfer sufficient information to ou
 tput. Unlike the supervised learning\, the information flow of unsupervise
 d learning has not been explored much. In this talk\, I introduce the info
 rmation extraction of unsupervised learning by using a representative gene
 rative model\, deep belief networks. Then\, I show that the unsupervised l
 earning maximally extracts relevant information given a fixed information 
 compression for its internal representations. This relation between the in
 formation bottleneck theory and our theory reminds the duality between the
  rate distortion theory and channel capacity in information theory. \n\n[
 국문] Information bottleneck 이론은 지도학습과정에서 일어
 나는 신경망의 정보흐름을 정량화하는 이론이다. 이 이
 론에 따르면 신경망의 학습은 입력과 출력을 연결시켜
 주는 최소한의 정보만 전달하고 불필요한 정보는 최대
 한 압축하는 과정임을 보여주었다. 우리는 입력과 출력
 이 정의되는 지도학습과 달리 비지도학습과정에서 일어
 나는 신경망의 정보흐름에 대해 연구했다. 이 발표에서
 는 대표적인 생성모형인 딥빌리프네트워크를 통해서 비
 지도학습과정의 정보추출에 대해 소개하려고 한다. 우
 리는 딥빌리프네트워크가 주어진 해상도의 내적표현을 
 가질 때\, 내적표현의 빈도수에서 최대한의 정보를 추출
 한다는 사실을 발견하였다. Information bottleneck 이론과 우
 리 이론의 관계는 마치 정보이론의 rate distortion theory와 c
 hannel capacity 사이의 쌍대성을 떠올리게 한다.\n\nhttps://ssh
 ep.snu.ac.kr/event/107/contributions/36/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/36/
END:VEVENT
BEGIN:VEVENT
SUMMARY:AI-based Smart Molecular Design
DTSTART;VALUE=DATE-TIME:20191108T042000Z
DTEND;VALUE=DATE-TIME:20191108T044000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-37@sshep.snu.ac.kr
DESCRIPTION:Speakers: Woo Youn Kim (Chemistry\, KAIST)\nThe ultimate goal 
 of chemistry is to make new molecules with desired properties. It is chall
 enging because chemical space is very large and discrete with a wide varie
 ty of molecules. For example\, there are only 108 molecules synthesized as
  potential drug candidates\, but 1060 molecules are estimated to be existi
 ng. High-throughput virtual screening approach has attracted great attenti
 on but still requires large costs and time. In this talk\, we propose to u
 se a molecular generative model based on deep learning algorithm as an alt
 ernative. It is specialized in controlling multiple molecular properties s
 imultaneously\, embedding them in namely the latent space. As a proof of c
 oncept\, we will show that it can be used to generate a number of molecule
 s as drugs with specific properties. We also apply it to design of new mol
 ecules with promising binding energy for a specific target protein and use
  them as potential drug candidates that are not in the database.\n\nhttps:
 //sshep.snu.ac.kr/event/107/contributions/37/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/37/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Harmonic Data Analysis for Shape and Centrality
DTSTART;VALUE=DATE-TIME:20191108T014000Z
DTEND;VALUE=DATE-TIME:20191108T020000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-38@sshep.snu.ac.kr
DESCRIPTION:Speakers: Woong Kook (Seoul National University)\nHarmonic dat
 a analysis aims to provide topological and combinatorial summaries of data
  sets by representing them as simplicial complexes. Topological data analy
 sis\, which we shall review briefly\, initiated a topological approach and
  introduced shape of data as a new data scientific feature. Recently\, the
  need for simplicial complexes for data analysis arose again due to the em
 ergence of simplicial networks for modeling higher order relations among d
 ata points\, which requires both topological insight and combinatorial pre
 cision. In this talk\, we will present methods from topological combinator
 ics for refining data shape via harmonic cycles and computing network cent
 rality via simplicial effective conductance. Applications to medicine and 
 social networks will be presented. We will also describe recent experiment
 s in machine learning incorporating mathematical data summary.\n\nhttps://
 sshep.snu.ac.kr/event/107/contributions/38/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/38/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Quantum AI\, Quantum Machine Learning
DTSTART;VALUE=DATE-TIME:20191108T012000Z
DTEND;VALUE=DATE-TIME:20191108T014000Z
DTSTAMP;VALUE=DATE-TIME:20260723T064752Z
UID:indico-contribution-107-39@sshep.snu.ac.kr
DESCRIPTION:Speakers: Dongpyo Chi (Seoul National University)\nRecent deve
 lopment in quantum technology together with advances in quantum algorithms
  impacts the field of quantum AI and machine learning. Many quantum machin
 e learning algorithms and their applications to AI are being appeared. We 
 present some of these and also our work in the field.\n\nhttps://sshep.snu
 .ac.kr/event/107/contributions/39/
LOCATION:Seoul National University\, College of Natural Sciences\, Buildin
 g 25-1 국제회의실 (International Conference Hall)
URL:https://sshep.snu.ac.kr/event/107/contributions/39/
END:VEVENT
END:VCALENDAR