A School in Computational Science & Engineering Richard Fujimoto Chair, Computational Science and Engineering Division
Georgia Tech Colleges Architecture Computing Ivan Allen Management Engineering Sciences Planned Changes School of Computer Science School of Interactive Computing Division of Computational Science and Engineering School CSE Division Focus on computer based models of natural & engineered systems 9 faculty with primary appointment in CSE 4 with secondary appointment t in CSE 9 adjunct faculty 15 faculty with joint appointments Operational Changes None: CSE already operating like a school in terms of education programs, administration, processes (e.g., RPT),
Rationale Create a home for faculty and students focusing on core topics such as modeling and simulation, computational data analytics, high performance computing CSE discipline recognized internationally (e.g., SIAM) Need for CSE departments recognized by (for example) NSF-OCI Such a home gives Georgia Tech a competitive advantage (e.g., faculty, student recruiting) relative to universities such as Stanford, Cal Tech, UC-Berkeley, UT-Austin, Univ. of Illinois, Purdue, Creates an administrative structure to manage resources (e.g., people) and foster increased collaboration especially with units in the Colleges of Engineering and Sciences Georgia Tech is taking a leadership position by establishing Computational Science & Engineering as a school
Education Programs Multidisciplinary MS and PhD degree programs in Computational Science and Engineering (began Fall 2008) Jointly offered by three colleges: Computing, Sciences, and Engineering 20 PhD, 20 MS students enrolled (Nov 2009) First distance learning degree offered by CoC Computer Science undergraduate program (revised 2007) Jointly offer with Schools of Computer Science, Interactive e Computing, and CSE Thread in modeling and simulation CRUISE: Computing Research Undergraduate Intern Summer experience (began 2008) Women and minority outreach
High Performance Computing Source: Austin American Statesman David Bader ( 96 Univ. Maryland) Jeff Vetter (98 ( 98 GT) joint with ORNL George Biros ( 00 CMU), joint Biomed. Eng.
Massive Data: NSF/DHS FODAVA Program Duke Georgetown Maryland Michigan Michigan State (Foundations of Data and Visual Analytics) CMU Northwestern Cornell FODAVA Lead: Georgia Tech Penn State Virginia Tech UIUC Princeton FODAVA partner universities UI-Chicago UC Santa Cruz UC-Davis Stanford Purdue Haesun Park ( 87 Cornell) Extracting knowledge from massive data critically important Georgia Tech won award as lead institution in FODAVA program Interdisciplinary team spanning CoC, CoE, CoS
High Performance Junior Faculty Alex Gray Machine Learning Data Analytics ( 03 CMU) NSF Career Award Rich Vuduc High Performance Computing ( 04 UC-Berkeley) NSF Career Award DARPA CS Study Group Guy Lebanon Machine Learning Data Analytics ( 05 CMU) NSF Career Award
Current Research Funding (Value of Active Grants & Contracts including CSE faculty as a PI or Co-PI - Nov. 2009) $ Millions Over $29 Million total in active grants (10 CSE faculty) Many multidisciplinary awards Four grants of $1M or more
Concluding Remarks CSE Division founded on the vision that Computational Science & Engineering is a discipline that merits a home on university campuses The CSE Division i i has grown and matured by Creating new educational programs Establishing strong, multidisciplinary research programs in HPC, Massive Data, and Simulation Providing leadership in the broader community The School of Computational Science and Engineering is a natural next step that will continue Georgia Tech s leadership role in defining and establishing the CSE discipline
The best way to predict the future is to invent it. - Alan Kay, 1971
Backup Slides
About the Name Pro Computational Science & Engineering the recognized, well established name in the community CSE faculty strongly favor this name for the school http://www.siam.org/students/resources/report.php p p p Con Potential confusion Computer Science vs. Computer Engineering vs. Computational Science & Engineering Potential ABET issues Use of Engineering outside College of Engineering Solution Retain name Computational Science & Engineering for school Avoid use of CSE name in undergraduate programs Document & publicize i distinctions/relationships among Computer Science, Computer Engineering, Computational Science & Engineering Continued communication among CoC and CoE, CSE and ECE
CSE Faculty David Bader High Performance Computing ( 96 Univ. Maryland) Alberto Apostolico Bioinformatics, Pattern Matching ( 76 Univ. Salerno) joint Inter. Comp. Guy Lebanon Machine Learning Data Analytics ( 05 CMU) George Biros High Performance Computing ( 00 CMU) joint Biomed. Eng. Jeff Vetter High Performance Computing ( 98 GT) joint with ORNL Ken Brown Quantum Computing ( 04 UC-Berkeley) joint Chemistry Mark Borodovsky Bioinformatics ( 76, Moscow Inst. Phs&Tech) joint Biomed. Eng. Haesun Park Scientific Computing Data Analytics ( 87 Cornell) David Sherrill High Performance Computing ( 96 UGa) joint Chemistry Rich Vuduc High Performance Computing ( 04 UC-Berkeley) Richard Fujimoto Parallel/Distributed Simulation ( 83 UC-Berkeley) Alex Gray Machine Learning Data Analytics ( 03 CMU) Hongyuan Zha Scientific Computing Data Analytics ( 93 Stanford)
Computational Science & Engineering (CSE) CSE is a discipline devoted to the systematic study of computer- based models of natural and engineered systems. Nanomaterials Astrophysics Transportation Weather and climate Biology (drug design, cancer treatment, phylogeny, ) Computation CSE Mathematics Biomedical Aerospace Manufacturing 14 Interdisciplinary collaboration with science and engineering
Disciplinary Strength CSE Strategy (our hedgehog concept*) Establish excellence in core areas of the CSE discipline: high performance computing, data and visual analytics, embedded modeling and simulation Interdisciplinary Collaboration NOT by simply providing a service (e.g., programming) NOT by simply applying existing techniques to applications (even important ones) BY advancing the state-of-the-art t th tin computational ti methods to enable solution of real-world problems BY defining and shaping new problems in the context of important application domains *J. Collins, Good To Great, Harper Collins, 2001
High Performance Computing The multicore challenge/crisis Single processor performance improvements essentially stopped ~2005 From handhelds to supercomputers, p parallel computing has become necessary to exploit new hardware capabilities HPC algorithms, computational methods, software to enable Understanding largescale networks (transportation, energy, water, Internet, ) Advances in Medical imaging Blue Gene/L Roadrunner ASCI Red ASCI White ASCI Q Earth Simulator ASCI Red Storm ASC NASA Purple Columbia Fluid simulation (e.g., heart valve design) Challenge: Scalable algorithms and software for million core machines. CSE Faculty: Bader, Biros, Fujimoto, Vuduc
A tsunami of data Massive-Scale Data Analytics Widespread deployment of sensors, camera, wireless networks, Internet 2002: 22 exabytes (10 18 ) electronic information and recorded media 2006: 161 EB digital information created, captured, replicated 2010: About 988 EB (almost 1 ZB) new information Limit it is our ability to extract t knowledge and insight i from the many sources of data Challenge: algorithms that scale to massive, complex data sets Epidemiology CSE Faculty: Apostolico, Gray, Lebanon, Park, Zha, Borodovsky Astrophysics Bioinformatics Homeland Security Text Analysis Biometric Recognition Social Networks
Modeling & Simulation Interacting with the Real World Emerging Technologies - Sensors and computers have become ubiquitous cell phones, ipods, video cameras, - Wireless communication widely deployed (e.g., cellular, WiFi, sensor networks) Embedded modeling and simulation - Advance from capturing the state of the world to predicting future states - Computation models interacting with the real-world - On-line management and optimization of systems Ad Hoc Distributed Simulation Vehicle to vehicle communication Roadside to vehicle communication Instrumented t traffic signal controller In Vehicle Simulations example Sophisticated distributed computing systems on the road for transportation system management Distributed simulation Wireless network protocols
Computers are incredibly fast, accurate, and stupid. p y,, p Human beings are incredibly slow, inaccurate, and brilliant. Together they are powerful beyond imagination. Albert Einstein