Transcript Research @ SOC Vice-Dean (Research)

Research Overview of
the Department of Computer Science
Prof NG Hwee Tou
Vice Dean (Research), SOC
14 September 2015
2
Main Research Clusters
Computational
Biology
AI
Prog Lang & S/W
Engineering
Analytics
Healthcare
MDA-Funded Centers
Media
Sustainability
Database
Systems &
Networking
Security
Felicitous Computing
Advanced Robotics
Cloud Computing
University-level Centers
• Interactive and Digital Media Institute (IDMI)
• Inter-disciplinary (computing, engineering, and social
sciences) research institute in interactive and digital
media
• NRF-funded CREATE Centres
• SMART (Singapore-MIT Alliance for Research and
Technology)
• E2S2 (Energy and Environmental Sustainability Solutions
for Megacities): NUS and Shanghai Jiao Tong University,
China
• NUS (Suzhou) Research Institute
• 20,000 m2 building
• Xu Guo Qin (director), Tan Tiow Seng (deputy director)
3
Felicitous Computing Institute
• Institute Mission:
to realize the original ideals of ubiquitous computing
“The most profound technologies are those that disappear. ... [Computing will
become] an integral, invisible part of people’s lives ... [where] the computers
themselves ... vanish into the background.” [Mark Weiser, Scientific American, 1991]
•
Activity Recognition
 research on machine learning,
music analysis, context
awareness, activity inference
•
Infrastructure for the
Internet of Things
 middleware for seamless plugand-play discovery and
connection of devices, using
smartphones as intelligent
gateways
4
Advanced Robotics Center:
Towards Human-Centered Collaborative Robotics
5
NRF/MDA-Funded Centres
• China Singapore Institute of Digital Media (CSIDM)
–
–
Partner: Chinese Academy of Sciences, Institute of Automation (CASIA)
Goal: Human language technology to enable communication in multiple languages
• NUS-Tsinghua Extreme Search Center (NExT)
–
–
Partner: Tsinghua University, China
Goal: Solve challenges in indexing, integrating, fusing and searching for live data
generated by millions of sensors, online forums, blogs and mobile devices
• Centre of Social Media Innovations for Communities (COSMIC)
–
–
Partner: Indian Institute of Technology (IIT), Bombay, and Nanyang Technological
University, Singapore
Goal: Empower the next 100 million people through social media innovations that
improve the way they live, work and play
• Sensor-enhanced Social Media Center (SeSaMe)
–
–
6
Partner: Zhejiang University, China
Goal: To build social cyber-physical systems to help tackle societal problems
related to lifestyle and security
Research @ CSIDM (http://www.csidm.sg/)
• Language mediation – Break the language barrier to
enable communication between speakers of different
languages
• Research focus:
– Language translation systems
• Text translation
• Speech recognition
• Speech synthesis
– Language learning
• Grammatical error correction
• Pronunciation evaluation
– Multimedia-facilitated chat system
7
Research @ NExT (http://next.comp.nus.edu.sg/)
• To acquire, aggregate and organize
continuous streams of live and dynamic
data to realize a smart environment
Integrate photos and check-in venues
Active mobile social network
Distribution of topics in Orchard Road,
Singapore
8
Automated info organization
Research @ COSMIC
(http://cosmic.nus.edu.sg/)
• To empower people in the community
through social media innovations that
improve the way they live, work, and play
• Focus on healthcare, infrastructure,
agriculture
9
Scalable for widespread deployment
Workable with limited hardware
Useable by semi-literate people
Affordable to low-income people
TOP: Ready
and
well-served
Middle 3
Billion
Bottom 2 Billion
• Develop tools and services for low
resource settings that are
•
•
•
•
Top
1.5
Billion
Information
gathering apps
Sense making
apps
MOP: Ready
but not
well-served
BOP:
Not ready
Information
dispensing apps
SeSaMe (Sensor-enhanced Social Media) Centre
Applications
Cyber
World
Physical
World
Social
Lifestyle
FitSense
Photo
Recommendation
Information-on-the-Go
Readpeer
Social Gaming
Security & Analytics
OneSpace
Video Analytics
Cyber-Physical
Systems
Social World
Interface/Interaction
Vision: “To build
social cyberphysical systems
to help tackle
societal
problems related
to lifestyle and
security”
10
Interactive
Augmented Reality
freeWalker
Intelligence
Aesthetic
QR Codes
CamTweet
Dynamic Path
prediction
Social Interaction
Detection
CamTweet: Conceptual Idea
Saint Patrick’s Parade is happening here !
Parade: 0.7
crowd: 0.6
Traffic: 0.2
…
Parade: 0.2
crowd: 0.4
Traffic: 0.6
…
Twee ng Camera for Event Detec on
Privacy
TaggedMapReduce
Data Indexing
and Analytics
Generic Parallel
Inverted Index
Foundations
Multimedia
Processing
Localization
Saliency
PiLoc
Artificial
Intelligence
Active Sensing
NRF CRP Program: Cloud and Big Data
Enhancement of
existing systems
MapReduce+
1.
2.
3.
1.
proposing new
systems for
hybrid workload
new architecture
for new
hardware/applications
11
epiC
LogBase
2.
3.
1.
2.
3.
Benchmarking Hadoop performance factors (VLDB 2010).
Llama: a column-based data warehouse on Hadoop
(SIGMOD 11)
AQUA: a query optimizer for HIVE (ACM SOCC 2011,
TKDE 2011)
Cayley-graph based index (VLDB 2011) and Indexing
(SIGMOD2010, VLDB2010)
ES2: an elastic storage for OLAP and OLTP (ICDE 2011)
E3: an elastic execution engine (J of Information Processing,
WISE2011)
Log store on the Cloud (VLDB2012)
Adaptive query processing for log store (ongoing work)
Transactions on log store (ongoing work)
2008-
2008-
2011-
Energy and Environmental Sustainability
Solutions for Megacities (E2S2)
CS-A: Waste
management in
megacities for
sustainability and
energy recovery
CS-B: Challenges of
emerging
contaminants on
environmental
sustainability in
megacities
SP-2
CS-B
Complexity
Environment
Size
CS-A
SP-1
12
SP-1: Missionoriented system
model for megacity
sustainability
SP-2: Distributed
clouds: Peta-scale
urban sensing and
data management
NRF Project
Trustworthy Systems from Un-trusted Components
Lead PI:
Enhancing local
capabilities
PIs:
Education - New
courses, …
Vulnerability
Discovery
Agency
Collaboration
– DSTA, …
Verification
Binary
Hardening
Data
Protection
Research Outputs
13
Industry
Collaboration
ST, Symantec,
NEC, …
Business Intelligence
(NUS-IBM Centre for Business Analytics)
• To develop systems and capabilities to
leverage the data, knowledge and
information from businesses and
organizations for intelligent decision
making and competitive advantage
Quantitative and Economic Aspects
14
Technical Aspects
15
Artificial Intelligence Cluster
• Theory of Computing: Learning theory (learning in the limit); Recursion
theory (what can/cannot be computed); Complexity theory (how costly);
Automata theory.
• Uncertainty in AI: Robot planning and learning involving partially
observable Markov decision process; Parallel Gaussian process
regression for real time prediction with big data; Information gathering
agents with active sensing.
• Spatio-Temporal Data Mining: Discovery of lag patterns (e.g., stock
profile); Trajectory Analysis (e.g., hurricane path).
• Image & Text Mining: Document image binarization for text retrieval;
Video text detection; Handwritten text recognition.
• Adaptive Decision Analytics: Context-sensitive decision support;
Causal knowledge discovery; Surprise-triggered adaptation; Multimodal
medical data fusion.
16
Computational Biology Cluster
Overview
• Members
–
–
–
–
Limsoon Wong
Wing-Kin Sung
Hon Wai Leong
David Hsu
• Research
– ~30 papers annually
– ~10 keynote and
other invited talks
annually
• Education
– 3-5 PhD students
graduate annually
• Service
– Editorial board of
Biology Direct,
Scientific Report,
BMC Research
Notes, JBCB, TCBB
– PC of all leading
comp bio
conferences
17
Computational Biology Cluster
Recent Notable Results
• Network-based analysis of transcriptomic & proteomic
profiling data
– High cross-batch reproducibility when sample size is small
• Protein-complex reconstruction from PPI networks
– Cutting-edge results for three challenging scenarios
• Drug resistance in pathogens
– Innovative host-pathogen protein-interaction prediction
– Phylogeny-aware drug-resistance mutation inference
• 3D chromosome modeling
– Guarantees recovering correct structure in noise-free case
• Discovery of disease-associated mutations
– Decoded complex patterns of genomic rearrangement in
hepatocellular carcinoma
Database Cluster
General
Extensible
A: Feed-forward
MLP
CNN
Auto-Encoders
Layer
Updater
B: Undirected
C: Recurrent
Easy to use
Scalable
Start a job by configuring
• NeuralNet
• TrainOneBatch
• Updater
• ClusterTopology
Training time
Cluster size
TrainOne
Batch
19
Media Cluster
•
•
Faculty: Michael Brown, Alan Cheng, Chua Tat Seng, Kan Min Yen,
Mohan Kankanhalli, Leow Wee Kheng, Low Kok Lim, Ng Hwee Tou, Ng
Teck Khim, Ooi Wei Tsang, Sim Khe Chai, Terence Sim, Tan Tiow
Seng, Wang Ye, Yn Kang Kang, Zhao Shengdong, Roger Zimmermann
Areas:
Computer Graphics & Computational Geometry: GPU for computational
geometry, computer animation
Computer Vision: computational photography, face recognition, medical
imaging
Human Computer Interaction: digital sand animation on multitouch tabletop,
interactive textual design and manipulation
Multimedia Systems & Retrieval: Media search, social media analytics, multisensor computing, multimedia security
Natural Language Processing: grammar checking, machine translation,
semantic & discourse processing
Sound and Music Computing: music information retrieval, healthcare
applications
Speech Processing: noise robust ASR, deep learning for ASR
20
Programming Languages &
Software Engineering Cluster
•
•
Faculty: Chin Wei Ngan, Dong Jin Song, Martin Henz, Aquinas
Hobor, Joxan Jaffar, Khoo Siau Cheng, Abhik Roychoudhury,
David Rosenblum, Roland Yap
Areas:
Constraints (CP/CSP): MDD/FA/Grammar constraints, Heap Solvers,
Higher Order Consistencies, Generating Geometry Questions, String
Solvers for JavaScript
Program Analysis: Termination Analysis, Resource Analysis, Worst
Case Execution Time, Symbolic Execution, Security Vulnerabilities in
Android Apps
Software Engineering: Code Clones, Specification & Bug Signature
Mining, Scalable Code Analysis, Quantifiying Uncertainity in Software
Testing, Program Repair & Patching, Regression Testing, Debugging
Verification: Concurrency, Separation Logic, Verifiying Heap
Programs/Data Structures/Heap Safety, Probabilistic Verification of
Markov Chains and Markov Decision Processes, Timed
Protocols/Systems, Model Checking
21
Security Cluster
Programming
Languages, Cybersecurity Education
Norman Hugh Anderson
System Security
Many best paper awards ICECCS 14,
FSE09, Usenix Security 07
Involved in Bitblaze
Liang Zhenkai
Network Security,
Applied Cryptography
Grant from TDSI, MINDEF
Chang Ee-chien
Binary Analysis,
Testing, Fuzzing
Best paper awards – FSE09
Tool deployment – Jslice
ACM Distinguished Speaker
Abhik Roychoudhury
System Security, Data
Protection
Best paper award ICECCS 14, Deployed
past research to Google+ and Chrome.
Prateek Saxena
Systems
security, PL
Grant from T-labs, FSTD
Roland Yap
CS Department Security Cluster is currently leading several visible research activities
including the NRF project “Trustworthy Systems from Un-trusted Components”, and
the National Cyber-security Laboratory (NCL) for experimentation.
22
Systems & Networking Cluster
• Mobile sensing, Internet of Things (Chan Mun Choon)
• Algorithms for highly dynamic, unreliable and distributed
environments (Seth Gilbert)
• Networks and distributed systems (Ben Leong)
• Economics of the Internet, big data analytics and cloud
computing (Richard Ma)
• Modeling program execution behavior (Soo Yuen Jien)
• Unmanned drones (Colin Tan)
• Multi-agent and real-time traffic simulation (Gary Tan)
• Parallel and cloud computing (Teo Yong Meng)
• Embedded computing, real-time systems (Tulika Mitra)
• Approximate computing, GPU computing (Wong Weng Fai)
• Efficient algorithms and bounds for dynamic network (Yu
Haifeng)
Funding Sources
• Ministry of Education (funds basic research)
– Faculty Level ~ $1.6M
• AcRF Tier 1 (FRC): < $180K
– University Level
• Special Grants
– Start-up grant for new faculty: ~ $100K – $400K
– Young Investigator Award: ~ $300K – $500K
– National Level
• AcRF Tier 2 (ARC):  $500K
• AcRF Tier 3: $5 – 10M (Type A)
$10 – 25M (Type B)
23
Funding Sources
• A*STAR (also runs research institutes)
– Applied research (Started to support basic research
in recent years)
– ~$100K – $1M
• Defence Organizations
– Ministry of Defence
– Defence Science & Technology Agency (DSTA)
– DSO National Laboratories (DSO)
– Oriented towards defence applications
24
Funding Sources
• NRF (National Research Foundation)
– Competitive Research Program ~ $10M
– Campus for Research Excellence And Technological
Enterprise (CREATE)
• Typically > $20M
– Initiative to develop digital media industry (~$500M)
• Administered by Media Development Authority (MDA)
• Typical funding:
– ~ 5 years
– ~ $10M
25
Working with Industry - Research
• Consulting
• Technology licensing
• Research contract
• Joint research labs /
sponsored research
26
• Examples of industry
sponsored research
– Fuji Xerox
– Google
– Huawei
– Intel
– Microsoft
Research Funding
• Annual new funding (S$ million)
AcRF
A*STAR
NRF
Others*
Total
FY2014
3.9
0.0
6.3
1.4
11.6
FY2013
2.2
0.6
0
1.7
4.5
FY2012
5.2
1
10.9
1.2
18.3
FY2011
3.6
1.7
1.3
3.9
10.5
FY2010
3.6
2.4
10
2.1
18.1
Note: *AOARD, DSO, Exploit Technologies, Fuji Xerox, Google, Huawei, Intel, Microsoft,
ODPRT, SERI, SMART, Symantec
27
Research Funding
• Annual expenditure (S$ million)
28
AcRF
A*STAR
NRF
Others
Total
FY2014
3.5
0.8
3.3
1.8
9.4
FY2013
3.5
1.3
4.1
1.8
10.7
FY2012
3.5
1
3.6
1.5
9.6
FY2011
3.2
0.9
2.7
1.1
7.9
FY2010
3.6
0.7
1.4
1.5
7.2
SOC Incubation Centre
Graduated
companies:
• BestPeer
• Cicada
• Decision-Ware
• Drive SG
• FlickEvents
• FriarTuck
• mChron
• MokoMomo
• Mozat
• PrivyLink
• SGEye
• Smoov
• Social@Work
• tenCube
• Thothe
• Tiny Whale
• Visenze
•…
29
Current incubatees:
• 6ESTATES Pte Ltd
• Accufind Technologies
• Collappe
• eBeeCare
• Investing Note
• JidoBox
• Logtomation
• OhMyEvent
• Stylr
Innovation & Entrepreneurship
Mr Wu Wenxiang graduated from
NUS School of Computing Class of
2007. His company, where he is one
of the co-founders, Zopim
Technologies Pte Ltd, has been
acquired by Zendesk, a cloudbased customer service software
solution provider, for US$29.8
million (S$37.3 million), as cited in
various media reports.
30