PDCC Seminar Series

1
PDCC Seminar Series

• Research Students

2
Objectives

• Discussion between groups, understanding what is
  in PDCC
• Students/Research staff to help each other
• Sharing of ideas/new research discoveries
• Platform to practice and improve presentations

3
Structure

• We have 2 initial seminars for research students
  and research staff (24th September)
• Divide the centre into 4 research area teams
• Seminar is once every 2 weeks
• Each week a team takes its turn with 1 person
  from the team presenting at that seminar.
• May have guest speakers when possible.
• Ideas and help of course are welcome!

4
What to talk about?

• Summary of your research.
• A new technology to share.
• Summary of a paper or a literature review.
• PhD transfer/viva practice.
• Tutorial on specific topic/software.
• Practice for conference paper presentation.
• .

5
TITLE OF RESEARCH NAME POSITION SUPERVISOR
PROJECT EMAIL

• Summary of Research
• What you topic of research is
• What you have done so far
• Current work
• Results
• Pictures
• Etc.

RESEARCH KEYWORDS
Areas of expertise What you can help people
with
Areas of you want/need to learn. What people can
help you with
6
Distributed Infrastructure
7
Optimal Resource Management Framework under
Uncertainty for Cloud Computing
Environments Sivadon Chaisiri (BOOM) PhD
student siva0020_at_ntu.edu.sg
Cloud computing, Green computing, Virtualization,
Optimization, Capacity planning

• Cloud computing
• Virtualization
• Parallel/distributed computing
• Deterministic/stochastic mixed integer
  programming

• Explore math models to optimally manage resources
  in the cloud under future price and demand
  uncertainty
• Contribution optimal virtual machine placement
  algorithms, and optimal power management for
  server farms
• Future work biologically inspired virtual
  machine placement, optimal capacity planning, and
  cloud market modelling

• Game theory (e.g., Nash equilibrium and supply
  function equilibrium)
• Biologically inspired computing
• Cost-benefit analysis
• Mostly require historical data from a commercial
  datacenter

8
Efficient and Robust HLA-based Simulations over
the Grid Supervisor Cai Wentong Li Zengxiang
Ph.D Student lize0001_at_ntu.edu.sg
Efficient and Robust HLA-based Simulations
Federate Migration
Federate
Efficient HLA-based Simulations
Federate Replication
DRC Migration
RTI
Project Objectives
Performance Enhancement, Fault Tolerance,
Distributed Simulation
Federate Replication
Federate
Robust HLA-based Simulations
Federate Recovery
- Parallel and Distributed Simulation -Synchroniza
tion -Migration Load Balancing -Fault
Tolerance -HLA -Grid
DRC Recovery
RTI

• Decoupled Federate Architecture
• Relay-based federate migration
• Replicating simulations with alternative
  synchronization approaches
• Checkpoint and sender-based message logging
  fault tolerance approach

-Load Balancing, Replication, Fault Tolerance in
Distributed System -Software Diversity -Interest
Management -Synchronization
9
A Service Oriented HLA RTI on the Grid (SOHR) /
DVE PAN Ke- PhD/PO student Sup Stephen John
Turner (PhD) Tang Xueyan
(PO) pank0001_at_ntu.edu.sg
SOHR
Distributed Simulation, HLA, Grid Computing,
Interest Management

• Distributed/Parallel Simulation
• Grid Computing, limited in GT
• Time Synchronization
• Interest Management (IM) in Networked Virtual
  Env (NVE)

• An HLA RTI implemented with GT4
• A Hybrid Time Synchronization algorithm based on
  both Con and Uncon Info
• An Efficient Sort-Based DDM matching algorithm
• A multi-user maze game on SOHR
• PO Work An hybrid IM mechanism for P2P NVE

• - Other aspects in NVE, e.g. QoS, consistency
  maintainance
• Online Gaming (Communication Architecture)
• Optimization

10
Data Grid Performance Analysis Zhang Junwei -
PhD Student Under A/P Lee Bu-Sung,
Francis jwzhang_at_ntu.edu.sg
Performance of Hierarchical Data Grid
Data Grid, Data Replication, Job Scheduling

• Data Grid System
• Data Replication Algorithm
• Job Scheduling Algorithm
• Performance Analysis

• Modeling Hierarchical Data Grid
• Predicting its optimal performance
• Analyze performance of replication algorithms
  with consistency management

• - Queuing Theory
• System Modeling

11
High Performance Computing NTU-IHPC Joint
Research Lab JIN Jiangming - PhD
student s080056_at_ntu.edu.sg
Mu
Performance Modeling for HPC application on
Multicore Architecture
Multi-Core Architecture, Performance Modeling
Multi-Core Architecture Auto-Tuning
Technology Component-based Parallel Processing

• Define Application and Architecture Models
• Formulate the Performance Metrics and propose
  optimization problem
• Solve given problem by efficient methods

Performance Metrics Modeling Applied Operations
Research Optimization
12
Distributed Systems
13
Consistency and Situation-aware Statue Update
Mechanisms in Distributed Virtual Environments
Supervisor - Prof Cai Wentong Li Yusen - PhD
student S080007_at_ntu.edu.sg
State Update, QoS, Consistency and
Situation-aware, Distributed Virtual Environments

• Focus on consistency issues due to network delay
  and resource limitation in client-server DVEs
• Investigate system model to define the QoS
  problem in formal style
• Explore status update polices to efficiently
  make use of limited resource for improving
  consistency in DVEs

Scalability and QoS issues in Distributed Systems

Consistency and QoS issues in other related
areas, e.g. database, web caching
14
Resource Provisioning in DVEs ZHANG Lu - PhD A/P
Tang Xueyan zh0007lu_at_ntu.edu.sg
This research aims to investigate
effective resource provisioning techniques for
distributed virtual environments. In a typical
configuration of distributed virtual
environments, entity states are maintained by a
group of servers. Participants, known as clients,
connect to the servers to send their actions and
receive updates. Thus, there are two types of
resources to provision servers and network
bandwidth. The research issues to be
investigated include analyzing appropriate
performance measures for resource provisioning
design resource provisioning techniques for
improving these measures joint optimization of
server provisioning and bandwidth provisioning.
DVEs, Resource Allocation, Client/Server
MMOGs, Distributed Simulation, Distributed
Networks
Distributed Networks Client/Server Architecture
Graph Theory Optimization Problem Simulation
15
P2P-based MMOG Liu Cheng PhD
student Supervised by Prof. Wentong
Cai LIUC0012_at_ntu.edu.sg
Peer-to-Peer, Massively Multiplayer Online Game,
Distributed Hash Table, Scalability, Reliability

• Efficient Infrastructure how to improve
  players experience in the game?
• Game World Partition Lookup
• Game State Update
• Scalable Infrastructure how to support
  hotspots in the game world adaptively?
• Workload Redistribution among Peers
• Reliable Infrastructure how to deal with lookup
  failure and missing game states?
• Node Failure Recovery

Structured Peer-to-Peer Network

• Network Simulation
• Networked Virtual Environment
• Optimization

16
Trusted Computing on Peer-to-peer Network Hao
Jianan- PhD student haojn_at_pmail.ntu.edu.sg
How to safely execute code on target device?
Foo(X), a
Foo(a)
Trusted Computing, Security, Peer-to-peer,
Virtual Machine, MMOG
Alice
Bob

• Challenges
• Can Alice know whether Bob faithfully execute the
  Foo?
• Can Alice get Foo(a) without explicitly telling
  Bob the value of a?
• Solutions
• Software-based approach
• Secure computation
• Hardware-based approach
• TPM (Trusted Platform Module)
• Virtual machine
• Dynamic root of trust

• Trusted Computing
• Cryptography
• Network Security
• P2P-based MMOG
• Computer Architecture
• Virtual Machine

• -Secure Computation
• Operating System
• Kernel Mode Debugging
• Cheating on MMOG
• Cryptanalysis

17
Data collection scheme in Wireless Sensor
Networks Zhao Wenbo- phd Candidate
zhao0101_at_ntu.edu.sg

• WSN Distinct characteristics
• -Egress traffic pattern
• -Sometimes mobile sensor nodes
• -Critical dependence on battery life
• Current work
• Interference free scheduling
• (TinyOS currently has a CSMA MAC,

Energy Efficient and Reliable Data Collection
-Adaptive data collection -Data consistency
maintenance -Energy efficient routing
scheme -TDMA/CSMA hybrid MAC
-Build up Mathematical Models -Algorithms -Simulat
ion
18

• Break time!

19
Efficient Wireless Sensor Data Aggregation
through Multi-path Routing Structures Luu Van Hai
PhD student Supervised by Dr. Tang Xueyan

• Summary of Research
• Research topic Wireless sensor data aggregation
  through multi-path routing structures.
• Challenges
• Constructing multi-path routing structures for
  robust data collection.
• Scheduling wireless sensors for efficient
  multi-path data aggregation.
• Current work
• Investigating the trade-offs among latency,
  energy consumption and robustness of multi-path
  data aggregation.

Keywords Wireless sensor network, multi-path
routing topology, data aggregation, wireless
scheduling.
I have some experiences in wireless sensor
network simulation, data aggregation methods,
algorithm analysis. Other people would help me
in approximate data collection, accurate data
aggregation.
20
Optimization
21
Operation Management in Container Terminals Guo
Xi- PhD student Supervisor Huang Shell
Ying guox0006_at_ntu.edu.sg
Operation Management/Optimization in Container
Terminals
Decision-Making, Optimization, Simulation

• Logistics related decision- making algorithms
• Dynamic Data-Driven Application Systems

• Real time management and optimization of
  material flow in container terminals
• Advanced dynamic dispatching algorithms and
  promised augmenting analysis prediction
  capabilities
• How to make decisions under tight time
  constraint and various uncertainties?

• Decision-making with uncertainty
• Mathematical Modeling
• Real-time Synchronization
• Robust Optimization
• -

22
Symbiotic Simulation Heiko Aydt - RA/PhD
student IMSS aydt_at_ntu.edu.sg
Automated Problem Solving in Symbiotic Simulation
Systems
Simulation-based Decision Support, Evolutionary
Computing

• Distributed/Parallel Simulation
• Evolutionary Algorithms
• Optimisation
• Agent-based Systems

• Symbiotic Simulation closely coupled simulation
  system and physical system
• What-if analysis process is simulation-based
  optimisation
• How to find an optimal solution to a given
  Problem?
• How to get around no-free-lunch (NFL) theorems?

• - Local Search Heuristics
• EA Parameter Control
• Showcases/Real-life Optimisation Problems

23
Truck Scheduling in Crossdocks
Scheduling Optimization Mojtaba Shakeri PhD
Student Malcolm yoke Hean Low X-docking
Planning Scheduling mojt0001_at_ntu.edu.sg
Logistics, X-docking, Scheduling, Optimization

• X-docking is a practice in logistics of unloading
  materials from an incoming truck and loading
  these materials directly into outbound trucks
  with little or no storage in between.
• Developing optimization algorithms for
  scheduling of trucks at x-docking terminals to
  minimize crossdocking total operation time
• Two dependency ranking heuristics (DR DR
  adaptive search DRAS) for sequencing queue of
  waiting trucks
• A machine fitness (MF) heuristic and dynamic
  programming (DP) approach for assigning sequenced
  trucks to crossdock door

- Search Methodologies in Combinatorial
Optimization (Constructive Heuristics) -
Operations Research (MIP, DP, BB) -
Deterministic Scheduling

• Local Search Heuristics
• Algorithm Complexity Analysis
• Just-in-Time (JIT) scheduling
• Stochastic Scheduling
• Parallel Optimization

24
A GENERIC BEE COLONY OPTIMIZATION FRAMEWORK FOR
COMBINATORIAL OPTIMIZATION PROBLEMS Wong Li Pei
(PhD stud) - wonglipei_at_pmail.ntu.edu.sg Dr. Low
Yoke Hean, Malcolm Dr. Chong Chin Soon

• Summary of Research
• Develop a generic framework based on the foraging
  behaviour of bees to solve combinatorial
  optimization problems such as TSP, ATSP, QAP and
  SOP.
• The framework has been tested on a set of TSP
  benchmark problems.
• Results shows that out of 84 problem instances,
  the BCO algorithm achieves lt1 from known optimum
  for 78 problems.
• Expanding the framework to cater for different
  types of problems.
• Enriching the framework with more intelligent
  behaviours.

Bee colony optimization, combinatorial
optimization problems, meta-heuristics.

• Population-based optimization techniques.
• Scheduling and timetabling.

• Statistical techniques for parameters control.
• Parallel computation.
• Population behaviours.
• Software engineering (generic framework creation).

25
Natural Applications
26
Pedestrian Simulation Hu Nan (PhD
Student) HUNA0002_at_ntu.edu.sg


3D Animation
Agent Simulation, Decision Making, Experience,
Pattern Matching

• Agent-based simulation of Pedestrians
• How does pedestrian behave in normal life
  situation?
• How to produce realistic navigational behaviors
  in model?
• Decision-making based on experience and pattern
  matching.

• Agent-based Model
• Navigational Behavior Simulation
• Pattern Matching

• Situation Definition
• Situation Classification
• Behavior Classification
• Behavioral Patterns

27
Large-Scale Agent-Based Simulation on Pandemics
LIM REAMSOVICHEA ream_at_pmail.ntu.edu.sg
Agent, Simulation, Pandemics, Multi-resolution,
Distributed System

• Development of algorithms and techniques for
  multi-resolution agent-based modelling
• Development of algorithms and techniques for data
  aggregation and integration into the simulation
• Development of a full-scale data-driven model of
  a particular pandemic case study

(I just started my PhD study)
Simulation Agent based system Distributed
System Pandemic models Multi-resolution
28
High Performance Computational Algorithms for DNA
Motif Finding Problems Sun Hequan- PhD
student sunh0013_at_ntu.edu.sg
Input sequences matrix tn (signal background)
ALGORITM (Cilk)
Motif signals
Bioinformatics, DNA Motif Finding, High
Performance Computing, Cilk
Combinatorics -Graph theory -Enumeration
Probability/Statistics -Expectation
Maximization, -Gibbs Sampling Parallel
Programming using Cilk

• Preliminary results an algorithm using EM and
  Median String Search is designed. It can give
  better information content while consuming less
  time (using Cilk) compared to MEME
• Currently we are working on weak motif finding
  and designing algorithms for constructing graphs
  and finding its cliques with size k by
  constructing trees from the graph using dynamic
  programming

Clique finding algorithms in graphs pattern
matching algorithms
29
Speech Reconstruction Hamid Sharifzadeh,
PhD Assoc. Prof. Ian McLoughlin hami0003_at_ntu.edu.s
g
Reconstruction of Natural Sounding Speech from
Whispers
Speech Processing, CELP Codec, Spectral
Enhancement

• Whispered Speech Characteristics
• Low Bit Rate Coding
• Speech Production
• LSP Modifications
• Spectral Enhancement
• Speech Analysis
• Image Processing

• Music Signal processing
• DSP Implementations of Music, Speech, and Audio
• Biomedical DSPs
• Noise Modelling
• Speech Recognition (HMM-Based)

30
Collaborative Computing
31
Collaborative Information Retrieval Mao Yuqing -
PhD student maoy0002_at_ntu.edu.sg Supervisor Prof.
Sun ChengZheng
Epistemology-based Social Search for Exploratory
Information Seeking
Social Search, Exploratory Information Seeking

• CSCW
• Human-Computer Interaction
• WWW
• Information Retrieval

• To improve exploratory information seeking by
  social search - utilizing the wisdom of crowds
• Epistemology is concerned with the knowledge
  derived from successful search processes
• Epistemologies are shared, reused, and refined
  by numerous users with same or similar search
  interests

• Semantic Web
• Social Network
• Security and Privacy

32
Real-time Collaborative Software Engineering with
Advanced Access Control Technology

• Incorporating Automatic Locking in Real-time
  Collaborative Software Engineering
• Achievement
• Real-time collaborative programming
• Simple access control functionality
• Current Work
• Automatic Locking

Fan Hongfei, PhD Student Supervisor Prof Sun
Chengzheng Email FANH0003_at_ntu.edu.sg

• KEYWORDS
• Real-time
• Collaborative editing
• Access control
• Automatic locking

AREAS Real-time collaborative editing Software
engineering Access control
33
Resolving Orthogonal Conflicts in Real-time
Collaborative Spreadsheet Applications Supervisor
- Prof Sun Chengzheng Wen Hongkai- PhD
student wenh0003_at_ntu.edu.sg
Concurrent operations might be generated from
arbitrary dimensions

• How to define the combined effects for
  orthogonal conflicting operations properly
• How to achieve the previous defined effects by
  designing correct transformation algorithms
• How to justify and formalize the combined
  effects and algorithms

Detection and resolution for orthogonal operation
conflicts
An extension for current Operational
Transformation(OT) Framework