[표지] 1
제출문 2
보고서 초록 3
요약문 4
SUMMARY 10
Contents 16
목차 17
제1장 연구개발과제의 개요 28
제1절 연구목적 및 필요성 28
1. 연구사업의 목적 28
2. 연구의 필요성 28
제2절 연구내용 및 범위 30
1. 계산과학공학 데이터 관리 및 처리 요소기술 개발 및 확보 30
2. 계산과학공학 플랫폼 가시화 기술의 수준 31
3. 플랫폼 활용커뮤니티 협력체계 활성화 32
제3절 추진전략 33
1. 연구사업 추진전략 및 방법 33
2. 대내외 추진체계도 34
제2장 국내외 기술개발 현황 분석 35
제1절 국내동향 35
1. 계산과학공학 데이터 관리 및 처리 요소기술 개발 및 확보 35
2. 플랫폼 적용을 위한 가시화 기술 동향 39
제2절 해외동향 40
1. 계산과학공학 데이터 관리 및 처리 요소기술 개발 및 확보 40
2. 플램폼 적용을 위한 가시화 기술 동향 72
제3장 연구개발 수행 내용 및 결과 77
제1절 계산과학공학 데이터 관리 및 처리 요소기술 개발 및 확보 77
1. 계산과학공학 데이터 플랫폼 요소기술 개발 77
2. 대용량 HPC 활용 AI 플랫폼 PoC 개발 124
제2절 계산과학공학 플랫폼을 위한 가시화 기술 개발 193
1. 하이브리드 렌더링 기술 개발 193
2. 플랫폼/가시화 서버 연동기술 개발 202
3. VR 가시화 기술 개발 204
4. 슈퍼컴퓨팅 응용을 위한 AI 활용 거대 컴퓨팅 기술 선행 연구 215
5. 슈퍼컴퓨터 5호기 활용 대표 응용 데이터 가시화 239
제3절 플랫폼 활용 계산과학공학 커뮤니티 협력체계 구축 및 활성화 251
1. 커뮤니티 협력체계 구축 및 활성화 251
2. 우수성과 홍보 (언론 보도) 254
제4절 지적 재산권 현황 256
1. 논문게재 실적 256
2. 학술발표 실적 256
3. 특허출원 실적 257
4. 특허등록 실적 257
5. 기술료 계약실적 257
6. S/W 등록 실적 258
7. 저작권(ISBN 등) 실적 258
제4장 목표달성도 및 관련 분야에의 기여도 259
제1절 목표달성도 259
제2절 관련 분야에의 기여도 261
1. 계산과학공학 데이터 관리 및 처리 요소기술 개발 및 확보 261
2. 계산과학공학 플랫폼을 위한 가시화 기술 개발 262
제5장 연구개발결과의 활용계획 263
제1절 계산과학공학 플랫폼 프로토타입 개발 및 전문 응용분야 적용 263
제2절 계산과학공학 플랫폼을 위한 가시화 기술 개발 264
제6장 참고문헌 265
[뒷표지] 271
〈Table I-1〉 Research Strategies and Methods 33
〈Table III-1〉 My Dashboard Management Module Spcification 80
〈Table III-2〉 Index Submission Management Module Spcification 86
〈Table III-3〉 Search Tool Module Specification 89
〈Table III-4〉 Advanced Kibana Search Module Specification 93
〈Table III-5〉 Advanced Search Module Specification 99
〈Table III-6〉 Comparison of RDS and VDI 111
〈Table III-7〉 Firewall Configuration to Deny All Access Except From Korea 112
〈Table III-8〉 PAM Configuration to Allow Two-factor-authentication 112
〈Table III-9〉 Firewall Configuration to Control Outgoint Access 112
〈Table III-10〉 Configuration(sshd_config) for Terminal-based Ssh Acceess Denial 113
〈Table III-11〉 Configuration of x2go for Server-side Clipboard Disable 114
〈Table III-12〉 PAM Configuration for User Jail 115
〈Table III-13〉 User Jail Creation Script 116
〈Table III-14〉 File Download Module Specification 119
〈Table III-15〉 File List Module Specification 122
〈Table III-16〉 Gradual Expanding Search Space Algorithm 126
〈Table III-17〉 The Hyperparameter space of XGBoost algorithm on UCI... 132
〈Table III-18〉 Average Time for all tasks on UCI regressions 132
〈Table III-19〉 AutoML Dataset upload module specifications 141
〈Table III-20〉 Data Processing code 148
〈Table III-21〉 Neural network code 151
〈Table III-22〉 Input Parameters 152
〈Table III-23〉 Usages of AI Studio API 152
〈Table III-24〉 MyStudio module specification 153
〈Table III-25〉 Workspace Module Specification 156
〈Table III-26〉 Dataset Module Specification 159
〈Table III-27〉 Jobs Module Specification 163
〈Table III-28〉 Example of Data Engineering Code 170
〈Table III-29〉 AI Network Designer Management Module Specification 175
〈Table III-30〉 Equation for calculating CNN 184
〈Table III-31〉 Source code for save My Favorite Network 190
〈Table III-32〉 rendering server funtions 194
〈Table III-33〉 Additional features for web visualization client 199
〈Table III-34〉 Server rendering test condition for web visualization 200
〈Table III-35〉 Test result of geometry transmission 201
〈Table III-36〉 Test result of server rendering performance 201
〈Table III-37〉 Pseudo code (atom generation) 206
〈Table III-38〉 Pseudo code (atom generation) 207
〈Table III-39〉 Operation list of vr controllers 209
〈Table III-40〉 Pseudo code for vr interaction 210
〈Table III-41〉 Additional function list for heterogeneous input devices 210
〈Table III-42〉 Pseudo code for session management 211
〈Table III-43〉 Pseudo code for device check 213
〈Table III-44〉 In-situ based data analysis techniques which are provided in the ALPINE for... 216
〈Table III-45〉 data size of millenium run simulation 239
〈Table III-46〉 AMR cell distribution of millenium run simulation (redshift 1.6) 239
〈Table III-47〉 Variable list of millenium run simulation (gas data) 242
〈Table III-48〉 Pseudo code for gas data load 243
〈Table III-49〉 Pseudo code for particle data load 244
〈Table III-50〉 An Example of meta file extraction 246
〈Table III-51〉 Pseudo code for vdb generation 247
〈Table III-52〉 Presentation titles in the the workshop of the 2020 KSEE conference 251
〈Table III-53〉 Detailed program in the the KSCFE 25th Short Course 252
〈Table III-54〉 Detailed program in the special session of the CDE 26th Conference 253
〈Table III-55〉 Media reports on computational science platforms 255
〈Table III-56〉 Thesis publication in journal 256
〈Table III-57〉 Presentation of journal 256
〈Table III-58〉 Patent application 257
〈Table III-59〉 Patent registration 257
〈Table III-60〉 Technical fee contract 257
〈Table III-61〉 S/W registration 258
〈Table III-62〉 Copyright 258
〈Table IV-1〉 Targeted goal achievement 259
〈Figure I-1〉 Relationship between sub-projects (R&D programs) 34
〈Figure II-1〉 HEMOS platform 37
〈Figure II-2〉 Concept of nano vfab 38
〈Figure II-3〉 Screen of iBat service 38
〈Figure II-4〉 A process of the Amazon Machine Learning Service 44
〈Figure II-5〉 Microsoft Azure Machine Learning Studio Overview 45
〈Figure II-6〉 AutoML services in the Google Cloud 46
〈Figure II-7〉 The IBM chatbot build service with the Watson assistant 47
〈Figure II-8〉 Recent competitions at the Kaggle 48
〈Figure II-9〉 An example of machine learning dataset at the OpenML portal 49
〈Figure II-10〉 CloudCV – Fabrik Platform 50
〈Figure II-11〉 Deep Neural Network in Sony's Neural Network Console 51
〈Figure II-12〉 Running training job in Sony's Neural Network Console 52
〈Figure II-13〉 Deep Congnition's Deep Learning Studio 53
〈Figure II-14〉 Data Preprocessing in Deep Congnition's Deep Learning Studio 53
〈Figure II-15〉 How H2O Framework Works 58
〈Figure II-16〉 Training Performance with Horovod 59
〈Figure II-17〉 Example using TensorFlow v1 with Horovod 60
〈Figure II-18〉 Amazon SageMaker: AutoML Service 61
〈Figure II-19〉 Google Cloud AutoML Service 62
〈Figure II-20〉 Screen of loading model with Netron 66
〈Figure II-21〉 Expresso's main screen 67
〈Figure II-22〉 Creating and modifying layers of a deep net 67
〈Figure II-23〉 The GUINNESS GUI 68
〈Figure II-24〉 network design in Barista 69
〈Figure II-25〉 Materials data search at the Materials Project portal 69
〈Figure II-26〉 SW statistics at NOMAD 70
〈Figure II-27〉 Approximated data size at the materials field 70
〈Figure II-28〉 NOMAD Encyclopedia 71
〈Figure II-29〉 Data analysis tutorials at NOMAD 71
〈Figure II-30〉 Web based Visualization Result at NanoHub 72
〈Figure II-31〉 Web based Visualization Result at NOMAD 73
〈Figure II-32〉 Visualization result using Deck.gl 74
〈Figure II-33〉 Visualization result using Kepler.gl 74
〈Figure II-34〉 Visualization result at CIMPLEX 75
〈Figure II-35〉 Visualization Result on the web using ParaViewWeb 76
〈Figure III-1〉 The architecture observation/experiment data platform 78
〈Figure III-2〉 My Dashboard Usecase Diagram 80
〈Figure III-3〉 My Dashboard Class Diagram 81
〈Figure III-4〉 My Dashboard Main Page 81
〈Figure III-5〉 Study Info Page 82
〈Figure III-6〉 Study Dataset Page 82
〈Figure III-7〉 Study dataset Info page 83
〈Figure III-8〉 Study Dataset Metadata Page 83
〈Figure III-9〉 Study Dataset Comment Page 84
〈Figure III-10〉 Study Detail Page 84
〈Figure III-11〉 Study Forum Page 85
〈Figure III-12〉 Study Forum Detail Page 85
〈Figure III-13〉 Index Submission Usecase Diagram 86
〈Figure III-14〉 Index Submission Class Diagram 86
〈Figure III-15〉 Index Submission Page 87
〈Figure III-16〉 Index Submission Study Choose page 87
〈Figure III-17〉 Data Management Process 88
〈Figure III-18〉 Search Tool Usecase Diagram 90
〈Figure III-19〉 Search Tool Class Diagram 90
〈Figure III-20〉 Search Configuration Page 91
〈Figure III-21〉 Searching Mode 92
〈Figure III-22〉 Advanced Kibana Search Usecase Diagram 93
〈Figure III-23〉 Advanced Kibana Search Class Diagram 94
〈Figure III-24〉 Advanced Kibana Search Main Page 94
〈Figure III-25〉 Advanced Kibana Search Configuration Page 94
〈Figure III-26〉 Advanced Kibana Search Statistics Page 95
〈Figure III-27〉 Advanced Kibana Search Discovery Page 95
〈Figure III-28〉 Advanced Kibana Search Configuration Page 96
〈Figure III-29〉 Advanced Kibana Search Configuration Target Page 96
〈Figure III-30〉 Advanced Kibana Search Index Search Target Page 96
〈Figure III-31〉 Advanced Kibana Search Study Select Page 97
〈Figure III-32〉 Advanced Search Usecase Diagram 97
〈Figure III-33〉 Advanced Search class Diagram 98
〈Figure III-34〉 Advanced Search Main Page 99
〈Figure III-35〉 Advanced Search Study Select 100
〈Figure III-36〉 Advanced Search Study Selected 100
〈Figure III-37〉 Advanced Search Field Select Page 100
〈Figure III-38〉 Advanced Search Query Operation 101
〈Figure III-39〉 Advanced Search Condition Selected Page 101
〈Figure III-40〉 Advanced Search Chart Result View Page 102
〈Figure III-41〉 Advanced Search Result List View Page 102
〈Figure III-42〉 Analysis Statistics View Page 103
〈Figure III-43〉 Sample Browser Main Page and Configuration 104
〈Figure III-44〉 Sample Browser Configuration 104
〈Figure III-45〉 Sample Browser Query Select 105
〈Figure III-46〉 Sample Browser Search Result 105
〈Figure III-47〉 Sample Browser Data Detail Page 106
〈Figure III-48〉 Sample Browser Aggregation View Page 106
〈Figure III-49〉 Search Tool Interface 107
〈Figure III-50〉 An Example of Variant Data Searching Result 107
〈Figure III-51〉 An Example of Detailed Vairant Data Searching Result 108
〈Figure III-52〉 An Example of Variant Data Filtered Search 108
〈Figure III-53〉 Aggregation of Variant Data Searching 109
〈Figure III-54〉 Secure Remote Desktop Environment Concept 110
〈Figure III-55〉 Login Screen of Secure Remote Desktop Environment 117
〈Figure III-56〉 OTP Authentication of Secure Remote Desktop Environment 117
〈Figure III-57〉 Secure Remote Desktop Environment 118
〈Figure III-58〉 Rstudio Application of Secure Remote Desktop Environment 118
〈Figure III-59〉 File Download Usecase Diagram 119
〈Figure III-60〉 File Download Main Page 120
〈Figure III-61〉 File Download Request Page 120
〈Figure III-62〉 File Download Request View Page 121
〈Figure III-63〉 File Download Individual Page 121
〈Figure III-64〉 File History Class Diagram 122
〈Figure III-65〉 File List Main Page 122
〈Figure III-66〉 File List Request Page 123
〈Figure III-67〉 Optimization with the Beale test function 128
〈Figure III-68〉 The Beale function optimization history plot 129
〈Figure III-69〉 The Hartmann 6-dimension function optimization history plot 130
〈Figure III-70〉 The Performance Evaluation for XGBoost comparing to existing HPO… 131
〈Figure III-71〉 Optimization history plot by time 135
〈Figure III-72〉 AutoML service on the PoC platform 140
〈Figure III-73〉 Usecase diagram: AutoML service 142
〈Figure III-74〉 HPO Job list in the portal 142
〈Figure III-75〉 An Example of the HPO Job View: Overview 143
〈Figure III-76〉 An Example of the HPO Job View: Trials Detail 144
〈Figure III-77〉 An Example of the HPO Job View: Resubmit 145
〈Figure III-78〉 An Example of the HPO Job on Jupyter 145
〈Figure III-79〉 The Concept of AI Studio, a web-based integrated development environment,… 147
〈Figure III-80〉 MyStudio Usecase Diagram 153
〈Figure III-81〉 MyStudio Class Diagram 154
〈Figure III-82〉 MyStudioJob Usecase Diagram 154
〈Figure III-83〉 MiniBbs Usecase Diagram 154
〈Figure III-84〉 MyStudio Detail View 155
〈Figure III-85〉 MyStudio New Workspace 155
〈Figure III-86〉 Workspace Usecase Diagram 156
〈Figure III-87〉 Workspace Class Diagram 157
〈Figure III-88〉 Dashboard Detail View 157
〈Figure III-89〉 Dataset Usecase Diagram 158
〈Figure III-90〉 Dataset Class Diagram 158
〈Figure III-91〉 Dataset Detail View 159
〈Figure III-92〉 Dataset CSV Detail View 160
〈Figure III-93〉 Dataset Image Detail View 160
〈Figure III-94〉 Dataset Others Detail View 160
〈Figure III-95〉 AI Studio's Overall Dataset Management View 161
〈Figure III-96〉 Notebooks Usecase Diagram 162
〈Figure III-97〉 Notebooks Detail View 162
〈Figure III-98〉 Jobs Usecase Diagram 163
〈Figure III-99〉 Jobs Class Diagram 164
〈Figure III-100〉 Jobs Training List 164
〈Figure III-101〉 Jobs Training Detail View 165
〈Figure III-102〉 List of Jobs for Data Engineering 165
〈Figure III-103〉 Job of Data Engineering Detail View 166
〈Figure III-104〉 Management of Training Job Detail View of AI Studio 167
〈Figure III-105〉 Management of Models Detail View of AI Studio 168
〈Figure III-106〉 AI Job Training Process and Monitoring through AI Studio API 169
〈Figure III-107〉 Data Engineering Job in AI Studio and Generated Result Table 171
〈Figure III-108〉 AI Network Designer scenario 175
〈Figure III-109〉 AI Network Designer Usecase Diagram 175
〈Figure III-110〉 AI Network Designer Class Diagram 176
〈Figure III-111〉 AI Network Designer Search 177
〈Figure III-112〉 AI Network Designer view mode 177
〈Figure III-113〉 Create AI Network Designer 178
〈Figure III-114〉 AI Network Designer creation popup 178
〈Figure III-115〉 Modify AI Network Designer 179
〈Figure III-116〉 Delete AI Network Designer 179
〈Figure III-117〉 AI Network Designer editor 180
〈Figure III-118〉 Top toolbar area 181
〈Figure III-119〉 Bottom toolbar area 181
〈Figure III-120〉 AI Network Designer Node area 182
〈Figure III-121〉 CNN Template 183
〈Figure III-122〉 LSTM Template 183
〈Figure III-123〉 Open template CNN and closed template LSTM 184
〈Figure III-124〉 Save and reuse My Favorite Network 185
〈Figure III-125〉 Delete My Favorite 186
〈Figure III-126〉 Neural network design 186
〈Figure III-127〉 AI Network Designer upload 187
〈Figure III-128〉 Pre-designed neural network upload result 187
〈Figure III-129〉 Download and export 188
〈Figure III-130〉 '.py' file converted to code and '.ipynb' file including API 189
〈Figure III-131〉 Neural network sharing screen 191
〈Figure III-132〉 Neural Network Sharing Popup 191
〈Figure III-133〉 Neural Network Sharing Results 192
〈Figure III-134〉 Neural Network Sharing Viewer 192
〈Figure III-135〉 An overall architecture of the hybrid server 195
〈Figure III-136〉 Components and data flows in the hybrid server 196
〈Figure III-137〉 A sequence diagram of the hybrid server 197
〈Figure III-138〉 Web visualization client according to rendering mode (left: full-size data,… 200
〈Figure III-139〉 The scenario of linking with the web portal workbench in the platform 202
〈Figure III-140〉 The visualization result of 1 tera bytes CFD simulation data… 203
〈Figure III-141〉 Event Scenario : WebXR based VR Visualization Client 204
〈Figure III-142〉 System Overview : WebXR based VR Visualization Client 205
〈Figure III-143〉 Atom rendering results 206
〈Figure III-144〉 Bonds rendering results 207
〈Figure III-145〉 Molecular data load and Lighting result 208
〈Figure III-146〉 Granting access to VR Device 213
〈Figure III-147〉 Stereo rendering results 214
〈Figure III-148〉 Rendering result of web-based vr molecular app 214
〈Figure III-149〉 Simulation and visualization resource configurations given the definitions… 215
〈Figure III-150〉 Comparison of the tightly coupled in-situ visualization and loosely coupled in-situ visualization 216
〈Figure III-151〉 ALPINE architecture 217
〈Figure III-152〉 InSituNet workflow 218
〈Figure III-153〉 PAVE architecture 218
〈Figure III-154〉 Vortex-Net architecture 219
〈Figure III-155〉 FlowNet architecture 220
〈Figure III-156〉 CNN based intelligent feature extraction & transfer function creation 221
〈Figure III-157〉 Overall architecture of the generative model for volume rendering 222
〈Figure III-158〉 High-resolution up-scaling for a low-resolution volume data using GAN… 223
〈Figure III-159〉 PointNet architecture 224
〈Figure III-160〉 MeshNet architecture 225
〈Figure III-161〉 Spirality of streamlines 226
〈Figure III-162〉 Comparison of applying and not applying scaling on the spirality function 227
〈Figure III-163〉 Curvature of the streamline 227
〈Figure III-164〉 LIC operation 228
〈Figure III-165〉 Overall procedure of applying LIC image as an input data into the… 228
〈Figure III-166〉 A Procedure of the quality improvement on a LIC image using the… 229
〈Figure III-167〉 Quality improvements of the 3D LIC volume 229
〈Figure III-168〉 Data augmentation of the LIC volume using randomly created white noises 230
〈Figure III-169〉 Applying a LIC image which generated from the input velocity fields… 231
〈Figure III-170〉 An architecture of the FCN based learning model... 231
〈Figure III-171〉 An architecture of the CNN based learning model predicting… 232
〈Figure III-172〉 Architecture of the deep learning model based on the dilated U-net 234
〈Figure III-173〉 Overall procedure of the proposed intelligent streamline visualization system 234
〈Figure III-174〉 User interface of the intelligent streamline visualization system 235
〈Figure III-175〉 Experimental results of the accuracy test varying the input feature data 236
〈Figure III-176〉 Experimental results of the accuracy test varying the quality of the… 237
〈Figure III-177〉 Experimental results of the accuracy test varying the number of filters… 238
〈Figure III-178〉 An example of voxel distribution in AMR volume data 240
〈Figure III-179〉 An example of overlapped block 241
〈Figure III-180〉 Bounding box of multiple block data (left) and rendering result (right) 241
〈Figure III-181〉 Voxel distribution in data block 242
〈Figure III-182〉 Voxel rendering result of VDB data 248
〈Figure III-183〉 Volume rendering result of VDB data 249
〈Figure III-184〉 Visualization result of VDB data 250
〈Figure III-185〉 Online presentation on the workshop of the 2020 KSEE con 252
〈Figure III-186〉 Presentation in the KSCFE 25th Short Course 253
〈Figure III-187〉 Presentation in the special session of the CDE 26th Conference 254
〈Figure V-1〉 Applicable plan of the visualization tool for computational science &... 264