Course Introduction Dr. Nawaporn Wisitpongphan
Dec 26, 2015
What do you need in order to conduct a PhD Research?A lot of reading in order to
Find a PhD thesis topicFind out what other have been done/doingFind out techniques which could be used in your thesis
Good Problem formulation (well-begun is half-done)Good techniques/tools
Data Collection & analysis
Good Results AnalysisGood Presentation *** (Journal Papers!!!)Self-Discipline & Self-LearningMotivation
What will we study in this class?Different types of Research
Theoretical StudySimulation StudyEmpirical Study
Some techniques about Data collectionsExperimental set upData Analysis toolsPresentation
Collaborative Learning. Learning by sharing. Nobody knows everything!!!
Goal of this classStudents obtain research skills necessary to conduct PhD-level research
Or better… Students obtains techniques that can be used to conduct their own research
Week Topics Remarks
1 (Nov 17)
Course Introduction + Research Topic Finding Class Activity
2 (Nov 24)
Students’ Presentation + Literature ReviewMini Research Project (Introduction)
Presentation
3 (Dec 1) Writing Technical Paper
4 (Dec 8) Reading Assignment (Data Clustering) No Class5 (Dec 15)
Data Clustering Techniques (Discussion)
6 (Dec 22)
(Project Proposal ) No Class
7 (Dec 29)
Take Home Exam (Writing Literature Review) No Class
8 (Jan 5) Literature Review Presentation (Exam Due) Presentation/HW#1
9 (Jan 12)
กี�ฬามหาวิ�ทยาลั�ย No Class
10 (Jan 19)
กี�ฬามหาวิ�ทยาลั�ย HW#1 Due No Class
11 (Jan 26)
Advanced Research Techniques (Optimization, Regression, Goodness of Fit, etc.)
HW#2
12 (Feb 2)
Advanced Research Techniques (Part II) HW#2 Due
13 (Feb 9)
Queuing Theory HW#3
14 (Feb 16)
Invited Talk (TBA) HW#3 Due
15 (Feb 23)
Project Presentation Presentation
16 (Mar 2)
Final Exam Paper Due
Class Info:References Grading CriteriaProbability, Random
Variables and Stochastic Processes by Athanasios Papoulis
Queuing Systems: Vol 1 by Leonard Kleinrock
Homework 30%Midterm 15%ResearchPresentation
10%Written Paper 15%Final Exam 30%
Class ProjectIndividual ResearchYou will get to
Analyze Huge amount of Data collected by NECTEC or Dek-D.com
Think about the research goal: What can you do with this data? What kind of information will be useful?
Generate some resultsWrite a paper about it : 5-6 page double
columnPresent it in the class
Class FormatClass Lecture: 1.5-2 hrsResearch Advising: 1-1.5 hr after the lecture
Class Presentation: 3 hrsGroup HW presentation: 0.5 hr
Assignment for next weekPresent your past research work1. Motivation & Goal2. Data Collection Technique3. Tools for Analyzing Data4. Your Potential PhD Topic
ExampleTitle:
Analysis of TCP and self-similarity of network traffic.
Motivation: Internet traffic is not Poisson!!! It’s heavy-
tailed.Data Collection Tool
OPNET simulatorStatistical/Analytical Tool
Variance-Time plotHurst Parameter
Self-Similarity: Proper DefinitionA stochastic process X(t) is statistically self-similar
with parameter H (0.5 H 1) if for any real a > 0, the process a-HX(at) has the same statistical properties as X(t)
This relationship may be expressed by the following conditions:
HX
X
H
H
a
asatRstR
a
atXVartXVar
a
atXEtXE
2
2
3
2
1
),(),(.
)]([)]([.
)]([)]([.
Degree of self-similarity: Hurst parameter (H) or self-similarity parameter is a measure of the persistence of a statistical phenomenon & is a measure of the long-range dependence of the stochastic processH = 0.5 Absence of self-similarityH = 1.0 Greatest degree of persistence or long-range dependence
The variance of the sample decreases more slowly than the reciprocal of the sample size
For most processes, the variance of a sample diminishes quite rapidly as the sample size is increased, and stabilizes soon
For self-similar processes, the variance decreases very slowly, even when the sample size grows quite large
Slowly Decaying Variance
Mathematically SpeakingThe m-average process of a
discrete-time stationary parent process X1, X2,Xn, … is
The variance is defined as
The variances of the aggregated process X(m) decrease linearly for large m
Hurst parameter,
...),,( )(3
)(2
)(1
)( mmmm XXXX
m
Njm
mji
mj jX
mX ,...,2,1
1
11
)(
m
N
j
mj
mN
m XXXVar1
2)()( 1
mXVarXVar m /)(
21 H
The ‘‘variance-time plot” is one means to test for the slowly decaying variance property
Plots the variance of the sample versus the sample size, on a log-log plot
For most processes, the result is a straight line with slope -1
For self-similar, the line is much flatterVariance-Time plot is obtained by plotting
log(Var[X(m)] )against log(m) where m is typically equal to 1 ms, 10 ms, 100 ms, 1 s, 100 s, …
Variance-Time Plot
Variance-Time PlotVariance-Time PlotV
aria
nce
m
Variance of sampleon a logarithmic scale
0.0001
0.001
10.0
0.01
100.0
Variance-Time PlotVariance-Time PlotV
aria
nce
m
Sample size mon a logarithmic scale
1 10 100 10 10 10 104 5 6 7
Data Collection TechniquesCollect throughput at the Network Bottleneck
Collect the throughput every 10 ms from OPNET Modeler
Use MATLAB to process data, e.g., Construct aggregated process, Find variance of aggregated process, plot V-T graph
Reference N. Wisitpongphan and J. M. Peha, “
Effect of TCP on Self-Similarity of Network Traffic ,” Proceedings of 12th IEEE International Conference on Computer Communications and Networks (ICCCN), Dallas, USA,Oct. 2003
STEP 1:How to find Research IdeasReading other papers (Lots of them) and write a summary which
answers some of these questions: From where did the author seem to draw the ideas? What exactly was accomplished by this piece of work? How does it seem to relate to other work in the field? What would be the reasonable next step to build upon this work? What ideas from related fields might be brought to bear upon this subject?
Exposing yourself to research Make a weekly effort to read at least the abstracts from good
journals in your field. Read 1-2 papers in detail. Attend a research seminar.
Directed Study Should you find the thesis advisor first or thesis topic first? Develop a thesis topic with your advisor using independent study
approach.
Rule of ThumbAvoid spending all your time doing literature
reviewGood research topic typically aims to solve a
certain problem (not EVERY PROBLEM) so.. Be clear on what your problem/motivation is Make legitimate assumptions on certain unknowns Be practical and realistic
Research = Work done by STUDENT not advisor “Student should know more than an advisor on the topic”
Do not wait for an advisor to tell you what to do, advisor’s job is to give student a feedback.
How to select a paper to read?First, find the major conferences in your
field.Find the conference programList the conference trackLook for a keyword e.g., name of the
technique, name of the problem, area of the study (typically listed in the title, abstract, or keyword of the paper)For example: VLDB 2012
http://www.vldb2012.org/general-information/advance-program/