Industrial Problems Seminar

AI and Industrial Mathematical Forum Guide (Daejeon Creative Economy Innovation Center)

Head of the Center Heo Jun | 2018-10-25 |Daejeon Creative Economy Innovation Center (KAIST 9F)

The event will involve about 100 participants from companies and start-ups in Daejeon, introducing research and industrial mathematics, troubleshooting cases,
It will consist of research and cross-enterprise networking.
Since this is the first time for business to be traded in Daejeon, we ask the department manager, PI, and interested researchers to participate.
The attached file is an event plan and I'll let you know once the program is finalized.
Please let me know if you have any questions.
□ Purpose
○ Introduce industrial mathematics concept and share cases such as artificial intelligence and big data analysis to solve corporate problems and increase sales
○ Activating a collaborative ecosystem between research institutes and engineers by matching companies interested in solving corporate problems using industrial mathematics such as artificial intelligence, big data analysis, and optimization
□ Period
○ October 25th 14:00-17:00
□ Location
○ Daejeon Creative Economy Innovation Center (KAIST 9F)
□ Subject presentation
○ Artificial intelligence based on industrial mathematics

Industrial Problems Seminar

Variable AutoEncoder and Related Works

Professor Choi Bong-dae | 2018-09-20 |Camp Middle Classroom

Variable AutoEncoder and Related Works
Professor Choi Bong-dae
Location: Camp Middle Classroom
Time: September 20, 2018 4:00 PM

Invitation seminar

Topological data analysis and its application

Kyung Sung Hyun Researcher | 2018-09-11 |National Mathematical Science Institute Mathematical Principle Application Center

Topological data analysis and its application
In the era of big data, supervised and unsupervised machine learning techniques were developed. These technologies
play a pivotal role in leading the artificial intelligence industry. In AI research, the prediction of new data is important
issue, and also identification (or definition) of subgroups is critical for the better results. In fact, the precise subgrouping
is important toward precision medicine. Topological data analysis (TDA) is the recently developed approach to identify
hidden subgroups based on the topology-based partial clustering technique. Mathematical background underlying TDA
and its application to the various research domains will be introduced during this seminar.

Industrial Problems Seminar

Analysis of Weather Information Using Deep Learning

Professor Han Seong-won | 2018-07-03 |Pangyo Start-Up Campus 1F Seminar Room 2

The Industrial Mathematical Innovation Center of the National Institute of Mathematical Sciences conducts various seminars with experts to find problems that can be solved with industrial mathematics, discuss solutions, and share successful cases through exchange of research with industry studies.
In response, I would like to invite Professor Han Seong-won of Korea University to conduct a seminar on industrial math and research exchanges. We ask for your interest and participation.
Title: Analysis of Weather Information Using Deep Learning
Speaker: Professor Han Seong-won (Koryo University)
Green: A study was conducted to apply artificial intelligence/dipning technology based on 4D space-time big data in search of weather journals. In this study, the characteristics of four-dimensional information, latitude, temperature, altitude, and time, were extracted using the deep learning technique, and the analysis of these characteristics provided a search system for similar days. In order to implement the search system with four-dimensional information, 10 layers of temperature and altitude data were used at the ECMWF analysis station, which was divided into 6 hours per day. In this study, we used a feature extraction technique applied to four-dimensional data with an Inception model, and we applied similar diaries to search algorithms. Using the pre-trained parameters obtained by mapping with ImageNet Dataset, diaries of the same date were also performed for 4D data, including altitude/temperature field/observe time, from Chapter 80 to GoogleNet model. Using these 80 features extracted, a 3-D sensor was created and a similarity calculation was performed. Similarity and similarity ranking were calculated using Mean Squared Error (MSE) among the features. In addition to the search for similar diaries using Deep Learning, this presentation also presented examples of the calculation of typhoon information using Deep Learning and the analysis of medical CT images.
1:00: 2018 7. 3 (Hwa) 16:00
Jang So : Pangyo Start-Up Campus 1F2
Participation application:icim.nims.re.kr
Question : 031-5171-5200 | info@icim.or.kr
※ All researchers, professors, students, business representatives, and ordinary people can attend.
※ If you would like to attend the seminar, please apply in advance on the Industrial Mathematics Portal(icim.nims.re.kr)
(This information is used only to promote events such as a seminar on Industrial Mathematical Innovation in the future.)

Industrial Problems Seminar

Dr. Kang: Max-Plank Institute for Plasma Physics | 2018-06-26 |Industrial Math Innovation Center on 4F of Start-up Campus 2 Please park in the basement and come up.

[ICIM] Seminar on Industrial Mathematical Exchange and Meeting (6.26)
The Industrial Mathematical Innovation Center of the National Institute of Mathematical Sciences conducts various seminars with experts to find problems that can be solved with industrial mathematics, discuss solutions, and share successful cases through exchange of research with industry studies.
I would like to invite Dr. Kang Gap-seok of the Max-Plank Institute for Plasma Physics to conduct a seminar and industry meeting.
1. Title: HPC in Europe and HLST
2. Dr. Kang: Max-Plank Institute for Plasma Physics
3. Pause/place: 26 June 2018 (Hwa)
- Seminar on Research and Exchange (11:00 - 12:00) / Seminar room on the 2nd floor of Pangyo Startup Campus
- Conference on Industrial Mathematics (13:00 - 14:30) / Industrial Math Innovation Center on 4F of Start-up Campus 2 Please park in the basement and come up.)
- Participation application:icim.nims.re.kr
- Door : 031-5171-5200 | info@icim.or.kr
※ All researchers, professors, students, business representatives, and ordinary people can attend.
※ If you would like to attend the seminar, please apply in advance on the Industrial Mathematics Portal(icim.nims.re.kr)
(This information is used only to promote events such as a seminar on Industrial Mathematical Innovation in the future.)
4. Content:
High Performance Computing (HPC) previously the domain of theoretical scientists and computer and software developers is becoming ever more important as a research tool in many surprising areas. The use of HPC in modeling complex physical phenomena such as weather, fluid dynamics, molecular interactions, astronomocal calculations and engineering design is well know to researchers in those fields. HPC is also now being used in industry to improve products, reduce production costs and decrease the time it takes to develop new products. As our ability to collect Big Data increases the need to be able to analyse the data also increases, this is an area HPC can be a most useful tool. More recently HPC is being used by researchers in social media, semantics, geology, archeology, materials, urban planning, graphics, genomics, brain imaging, economics, game design and even music.
The infrastruture and usage HPC in the Europe had underdeveloped in comparison with USA. From 2009, the investments on HPC of Europe expecially Germany have been expanded. The investments have included in the infrastructure and expand the usage through supporting scientists and training new researchers. The High Level Support Team (HLST) in the EUROFusion project in the Horizon2020 framework was consisted to support developing Fusion simualtion codes.
In this talk, I present the current status and plan of future investment of the HPC in Europe. I also talk about my experiences as a memeber of HLST.

Postdoc seminar

Model-based approach to load balancing for heterogeneous systems

추형석 (국가수리과학연구소) | 2015-01-12 16:30~17:30 |수학원리응용센터 중형세미나실

We investigated how to distribute computations among computational resources under data parallelism. A min–max model of computation times was proposed to reflect a heterogeneous computing system. Time functions for each resource were estimated with reference parameters, and sampling statistics evaluates those parameters such as effective memory bandwidths and FLOPS. Our min–max model includes those time functions as objective functions so that it suggests load balancing point for an arbitrary problem size. Several BLAS examples confirm that our model fits well comparing with real heterogeneous computing with OpenCL.

Industrial Problems Seminar

김진범 박사 (Barclays-New York, U.S.A) | 2014-10-06 16:30~17:30 |수학원리응용센터 중형세미나실

신용상품에는 크게 국가 및 회사은행개인의 신용을 거래하는 금융 상품으로 크게는 채권과 Credit Default Swap 및 옵션이
있습니다. 이 시간에는 첫번째로 각각의 상품의 개요와 미국 신용 상품 거래 시장이 어떻게 발전해 왔는지에 대해서 개략적
으로 설명하겠습니다. 두번째로는 2008년 금융 위기 이후에 새로 바뀐 제도적 규제가 신용상품 거래 시장에 어떠한 영향을
미쳤는 지를 설명하겠습니다. 마지막으로 이 규제로 인해 현재 은행 및 투자자들이 당면한 문제들과 어떻게 이것을 해결할
방법이 있는지에 대해서 알아 보는 시간을 가지겠습니다.

Industrial Problems Seminar

노건엽 박사 (보험개발원 계리 및 리스크 팀) | 2014-09-17 16:30~17:30 |수학원리응용센터 중형세미나실

보험상품은 대수의 법칙(the law of large numbers)이라는 확률론적 방법에 기반하여 만들어지는 금융상품으로 수리과학이
필수적인 역할을 합니다. 최근에는 금융공학 및 빅데이터 분석방법 등과 결합하여 다양한 적용 사례를 보여주고 있습니다.
보험상품에 내재된 옵션 평가, 현금흐름 평가를 위한 장기 이자율 시나리오 생성, 클러스터링을 이용한 보험계약 군집화 등
실제 사례를 통하여 보험에서 수리과학이 어떻게 활용되고 있는지 설명할 것입니다.
또한 보험산업에서 수리과학의 역할과 전망에 대해서 논의해 보려 합니다.

Industrial Problems Seminar

이원석 상무 (삼성전자 Software R&D Center) | 2014-08-28 11:00~12:00 |수학원리응용센터 중형세미나실

Superhyhobic micro-structured surfaces utilize a combination of chemical treatment and microscale surface
topography to increase the hyhobicity of a surface. Superhyhobic surfaces exhibit a number of interesting
properties such as dramatically decreased wettability and reduced friction in laminar flows. These characteristics make them a potential enabling technology for a variety of applications including microfluidics, lab-on-a-chip devices, self-cleaning surfaces, and reduction.
In this talk, we discuss the reduction capability of the microstructured surfaces to understand the fundamental
interfacial phenomena that allow these surfaces to exhibit such properties. Along with experimental efforts, numerical
simulations and theoretical studies have been conducted. Among those parallel efforts, in this seminar, we focus on the
analytic solution method using conformal mapping to solve pressure driven fluid flow problems under mixed boundary
conditions within microchannel.

Invitation seminar

Mechanization of Proof: From 4-Color Theorem to Compiler Verification

허충길 교수 (서울대학교 컴퓨터공학부 소프트웨어이론 연구실) | 2014-08-27 16:30~17:30 |수학원리응용센터 중형세미나실

I will give a broad introduction to how to mechanize mathematics (or proof), which will be mainly about the proof
assistant Coq. Mechanizing mathematics consists of (i) defining a set theory, (2) developing a tool that allows writing
definitions and proofs in the set theory, and (3) developing an independent proof checker that checks whether a given
proof is correct (ie, whether it is a valid combination of axioms and inference rules of the set theory). Such a system is
called proof assistant and Coq is one of the most popular ones.
In the first half of the talk, I will introduce applications of proof assistant, ranging from mechanized proof of 4-color
theorem to verification of an operating system. Also, I will talk about a project that I lead, which is to provide, using Coq,
a formally guaranteed way to completely detect all bugs from compilation results of the mainstream C compiler LLVM.
In the second half, I will discuss the set theory used in Coq, called Calculus of (Inductive and Coinductive) Construction.
It will give a very interesting view on set theory. For instance, in calculus of construction, the three apparently different
notions coincide: (i) sets and elements, (ii) propositions and proofs, and (iii) types and programs.
If time permits, I will also briefly discuss how Von Neumann Universes are handled in Coq and how Coq is used in
homotopy type theory, led by Fields medalist Vladimir Voevodsky.