Transcript Slide 1
Week 5. Writing the PhD thesis,
proposal and research papers
PhD seminar
Dr. Felipe Orihuela-Espina
Outline
1.
2.
3.
4.
How to elaborate a PhD thesis
The thesis protocol
Parts of a thesis
Writing abstracts
18/07/2015
INAOE
2
HOW TO ELABORATE A PHD
THESIS
18/07/2015
INAOE
3
How to elaborate a thesis
Definition: What is a thesis?
What has a thesis got to demonstrate?
How to choose a thesis topic
The actors
The student
The supervisor
The thesis committee
The panel
Developing the project
Periodic monitoring by the supervisor and the committee
Writing the document
Formatting
How to write the thesis
Legal authorship (Institution) and intellectual authorship
(student)
18/07/2015
INAOE
4
Some references for this section
[Camacho2003] Camacho Mejía, Felipe;
Herrera Barrera, Armando; Guía para la
elaboración de una tesis (2003)
Universidad Autónoma del Estado de
México (UAEMex)
[Sloman??] Sloman, Aaron “Writing a
thesis” University of Birmingham
[??] “How to write a PhD thesis”
18/07/2015
Dr. Felipe Orihuela Espina
5
Definition of a thesis
A thesis is a scientific text describing a an
in-depth description of a phenomenon
(that includes computing) together with a
set of solutions/hypothesis to yet unsolved
problems or questions about the
phenomena, and that further provides
extensive support and evidence to back up
any claims made. [Self definition]
18/07/2015
Dr. Felipe Orihuela Espina
6
What has a thesis got to demonstrate?
That the student is capable of:
Carrying out innovative research by himself
Criticism; both towards other’s work and to his
own work
Planning, executing and finishing a long term
project
Communicating knowledge and ideas
…in an organised fashion understandable by nonexperts
…yet without sacrificing accuracy
18/07/2015
Dr. Felipe Orihuela Espina
7
How to choose a thesis topic
Proposed by the department or the supervisor
Often as a result of a research project, a
necessity/research priority of the institution or sometimes
from a collaboration with an industrial partner
Pros:
Saves some headaches such as questioning one-self what do I
do?
The problem definition, extension and coverage are often well
studied and defined
Motivation and justification is clear
There is already a researcher interested or involved, almost
certainly with expertise on the topic
Cons:
You might not love the topic
You might not get on well with the supervisor
You risk just following instructions, and not demonstrating
initiative (necessary for obtaining the degree)
18/07/2015
Dr. Felipe Orihuela Espina
8
How to choose a thesis topic
Proposed by the student
Often as a result of the student’s interests
Pros:
You are likely to love the topic
You are free to find a supervisor which you want to
work with
You are already demostrating drive and initiative
Cons:
There might not be any researcher willing to supervise
the thesis or with enough knowledge about the topic
You’ll have to work on the problem definition,
motivation, etc
Funding?
18/07/2015
Dr. Felipe Orihuela Espina
9
The actors
The student
The supervisor
The thesis committee
The panel
18/07/2015
Dr. Felipe Orihuela Espina
10
The student
You! The one carrying out the thesis project
Responsible for:
Solving all problems related to the thesis research
Carry out the experiments
Proposing the solutions
Writing and submitting the protocol, the thesis and any papers that may
result
Successful finalization of the research
If something goes wrong it is your fault, so do not blame your supervisor.
The scientific quality of the work (legal)
Finding funding for his/her conference attendance
Entitled to:
Institutional support
An adequate supervision
That’s not mean babysitting
18/07/2015
Dr. Felipe Orihuela Espina
11
The supervisor
Whether the one who originally offered the topic or
chosen by the student.
That’s not your boss, just your companion!
In every aspect of the research, you have the last word.
It’s your thesis, not his/hers.
Responsible for:
Providing adequate guidance all throughout
Providing starting knowledge/context about the problem
The scientific quality of the work (moral)
Ensuring that the student finishes in time and form
…despite the student’s best efforts not to…
Entitled to:
When appropriate, co-authorships in publications
18/07/2015
Dr. Felipe Orihuela Espina
12
The supervisor
NOT responsible for solving anything at all about the
thesis.
The student shall not assume that his supervisor knows
everything about the topic
Moreover, after a year or so the student ought to have
surpass his supervisor in knowledge about the topic.
Often can recommend bibliography and references
May help to a degree in the writing and proof reading
of the thesis
May suggest a publication strategy
May help in funding search for conference attendance
18/07/2015
Dr. Felipe Orihuela Espina
13
The supervisor
The supervisor is responsible for the
scientific quality of the work, as in time it
will become part of his CV and
consequently linked to his/her reputation
However
This responsibility is only moral, not legal.
The student is free to submit his work even
without the consent of the supervisor and it is
not obliged to comply with any of the
supervisor recommendations.
18/07/2015
Dr. Felipe Orihuela Espina
14
The thesis committee
4 members of the faculty plus 1 external
Not all will have expertise related to the thesis topic, this
offers a different point of view
Ultimately chosen by the faculty,
…but the student and (perhaps) the supervisor may
proposed alternatives
Responsible for:
Periodically evaluating the student progress
Detecting weaknesses in the research
Providing suggestions for amending deviations menacing
the finalization within time constraints
Entitled to:
Being provided with the advances with enough time
…in English if Spanish is not their first language!
18/07/2015
Dr. Felipe Orihuela Espina
15
The panel
Peers assessing the final workpiece
Experts on the field
Responsible for:
Evaluating the final thesis
Rejecting works which do not exhibit originality, or
do not reach scientific standards
Entitled to:
Being provided with the advances with enough
time
…in English if Spanish is not their first language!
18/07/2015
Dr. Felipe Orihuela Espina
16
How to write the thesis
Developing the project
Periodic monitoring by the supervisor and
the committee
Writing the document
Formatting
18/07/2015
Dr. Felipe Orihuela Espina
17
Developing the project
A coarse overview:
Reading about the background
Establish a calendar
Tentative: if no special constraints applied. Admit deviations
Imposed: when special project constraints applied. Does not
admit deviations.
If in engineering; then budget proposal
18/07/2015
Inc. materials and human resources
Launch hypothesis and analyse project requisites and
demands
Design and execute experiments
Analyse your data
Write your final thesis document
Dr. Felipe Orihuela Espina
18
Periodic monitoring
The goal is to ensure that the thesis is finished in
due time and form guaranteeing minimal scientific
quality (and engineering if it is the case)
Advances should be check periodically
The most common way for this monitoring are the
bi/weekly meetings with the supervisor
Thesis committee meetings often take place every 6
months and are exceptional occassions to assess the
real progress of the thesis
Other forms of monitoring include:
Seminar giving
Technical reports writing
Periodic report writing as requested by the
programme
18/07/2015
Dr. Felipe Orihuela Espina
19
Periodic monitoring
How often should I meet my supervisor?:
It really depends on your needs…
However:
Do NOT allow your supervisor to avoid meeting
you for long periods
Do NOT relax yourself and “forget” to meet you
supervisor for long periods
18/07/2015
Dr. Felipe Orihuela Espina
20
Periodic monitoring
What should I take to/prepare the meetings?
Read in depth about the specific topic to discuss that week
Ensure that you have your questions ready
Organize any results (figures and tables) that you will show
It is rude to do it at the expense of the time of your supervisor
If you have become stuck onto a particular problem
Do not tell your supervisor until you have at least seriously tried to solve it
…but if really stuck, then spend the most of the meeting on this.
Remember that your supervisor may not know the solution!
Have a list of non-research issues that ought to be disscussed: admin,
conference attendance, scholarship problems, etc
Do not expect your supervisor to remember from one meeting to the next you
current needs
If presenting a document (paper, report, protocol, etc) be sure that the
draft is polished to your best
It is rude to use your supervisor as a spellchecker or as a primary school teacher
who has to tell you that every sentence you write is correct
Respect the time of your supervisor!
…during the meeting, and out of the meeting
If exceptionally arriving late then apologize in advance by mail/phone
Do not go beyond your scheduled time; perhaps other students are waiting
18/07/2015
Dr. Felipe Orihuela Espina
21
Periodic monitoring
Your supervisor should in turn:
Read whatever you sent prior to the meeting
BEFORE the meeting
Be up-to-date with your research
Not impose his will or point of view; but
instead give his/her best advice and let the
student take the decision
Tell the student if other meetings and
responsibilities comes in the way
Reallocate time if necessary so that the student do
not miss his meeting
18/07/2015
Dr. Felipe Orihuela Espina
22
Writing the document
Who is it written for?
Scientific peers
The thesis is a scientific document:
Be precise and concise
Effort to ensure efficient and effective communication
Avoid colloquial laguage
Be scrupulous on correctness (lexical, orthographical, syntax,
grammar, etc)
Artistics licences are valid as long as they do not prevent
clear communication of ideas
Long sentences are often a bad idea
Take care of the flow of ideas
The reader does not know:
your nomenclature, nor your acronyms, etc
the details of your research and specifically of your
experimentation. Provide enough details for replicability.
18/07/2015
Dr. Felipe Orihuela Espina
23
Writing the document
Assume knowledge only within expectable limits
Level of detail
Basic concepts
If truly trivial, then omit them
…but remember that the reader knowledge does not match
exactly yours!
Things which are trivial for you, may not be trivial for him.
Advanced concepts
If fairly known, mention and provide adequate references.
Concepts developed in your thesis
They ought to be perfectly/accurately defined, and if
necessary with proper mathematical formulation.
Avoid explining then in several places
…however readability may demand some repetition
18/07/2015
Dr. Felipe Orihuela Espina
24
Writing the document
Level of details
Experimental description
The secret is simple: Replicability
Every aspect of your experiment has to be replicable with absolute
fidelity by the reader.
Not replicable from your text, not good enough
Often convenient to provide justification for arbitrary decisions.
Mathematical development, proofs and
demonstrations
Omit if they are trivial (although they almost never are)
Explain them as they unfold; do not only state them
A thesis is not better or more complex conceptually just because
it has more maths and looks like more formal
…that may scare a bachelor but not your panelist, and in turn it may
actually irritate them
Ensure that they are utterly correct and coherent thorughout the
18/07/2015
document
Appendix are a good place for long developments and proofs
Dr. Felipe Orihuela Espina
25
Formatting
Compulsory: Most times, the institutions have
regulations about the formatting (inc. front
cover, style, fonts, interline space, margins,
paper size (and weight!), etc). Be sure you
comply with them.
Availability; We may not have always the
software we like. Use whatever you need to
guarantee the best possible presentation.
Never excuse a bad presentation blaming the
software/hardware tool you have used!
Portability: The format chosen has to be
easily portable to other formats.
18/07/2015
Dr. Felipe Orihuela Espina
26
Authorship
Legal authorship (Institution)
The institution is the legal responsible for the
results.
The institution can legally protect the intellectual
property of the knowledge generated by the
research and/or commercially exploit the
research
…but in turn any damage that your research may
cause is the institution’s responsibility
Silly example: If the bridge falls, its not your fault, but the
institution!
Any other (internal) researcher can continue your
work (e.g. use your data!) without asking for
permission
Although it is considered polite to ask for it…
18/07/2015
Dr. Felipe Orihuela Espina
27
Authorship
Intellectual authorship (student)
As the original author of the thesis, the student
keeps the intellectual property
The student is therefore entitled to (and in fact
more often than not, encouraged to) publish your
work
After published, any other (external) researcher can
continue your work without asking for permission
The student can self-cite himself without
committing plagiarism (but beware of selfplagiarism and salami slicing issues!)
The student can continue its research on the
topic even after finishing its liase with the
institution.
18/07/2015
Dr. Felipe Orihuela Espina
28
THE THESIS PROTOCOL
18/07/2015
INAOE
29
The thesis protocol
Title
Abstract
Introduction and related work
Justification and Motivation
Defining limits and assessment
The research questions
Goals; Main and specific
Theoretical and reference frameworks (literature review)
Methodology; Materials and Methods
Research hypothesis statement
Planning and description of tasks and experiments
Scheduling
Publications derived from this thesis (Optional)
Conclusions
References
18/07/2015
INAOE
30
Title
It is convinient to set a working title early in
your PhD
This might not necessarily be the one at the
end of your thesis (in fact, it will likely not
coincide)
It helps centering the thesis main topic
Try to be as concise as posible to delimit the
phenomenon
A rather generic title may give a wrong impression
to the thesis committee
18/07/2015
Dr. Felipe Orihuela Espina
31
Title
Example:
Orihuela-Espina PhD Thesis
Working title: “Interpretation of fundus images
using physics based models.”
Final (Real) thesis title: “Modelling and verification
of the diffuse reflectance of the ocular fundus”.
18/07/2015
Dr. Felipe Orihuela Espina
32
Abstract
It has to include:
The purpose or specific aim of the project
The methodology or experimentation
procedure
The expected results
The main expected conclusion and impact
Writing abstracts is an art: we will learn
more later
18/07/2015
Dr. Felipe Orihuela Espina
33
Abstract
Examples:
Prahl89 - PhD Thesis – Example for a thesis
Berenschot2002, Costa2000, DeLint2000 –
Structured abstracts
Cheong90 - Review, therefore it lacks results;
A fantastic example of short and concise
Edelsbrunner83 – Short and concise
18/07/2015
Dr. Felipe Orihuela Espina
34
Introduction and related work
It should describe some relevant prior
knowledge
It should very briefly summarised the state of
the art
A rather naive way of doing it is by enumeration:
“This fellow did this. This fellow did that, …”
You’ll be shouting that you’re a novice in the field
A more elegant/smart way is to tell the story with
a good flow of ideas, and simply drop the key
references where suitable
In computer science, related work is often
written in a separate section.
18/07/2015
Dr. Felipe Orihuela Espina
35
Justification and Motivation
You do not take over a PhD because you are
bored at home…
…but because there is a real need to
understand a phenomena
Economical
Scientific
Academic
Others
Nope! “…because I need it to get my degree” is
NOT a valid justification even if it is the only real
one
18/07/2015
Dr. Felipe Orihuela Espina
36
Justification and Motivation
Research hypothesis statement
It demands a good analysis of the requisites.
They may change as you progress, but the
more honest way to proceed is to keep it fixed
and if necessary when writing your thesis
state both, the original hypothesis and the
new one explaining why the change of mind
18/07/2015
INAOE
37
Defining limits and assessment
Clearly indicate the limits of your work
The PhD is not the moment to get a Nobel prize
Try to be realistic, it is easy to under-/overestimate your capacity
It is at times hard to differentiate limits from
your goal
If too short; maybe not enough to get the degree
If too ambituous; maybe unfeasible and unable to
defend
18/07/2015
Dr. Felipe Orihuela Espina
38
Defining limits and assessment
Limits are often stated in terms of what’s
goind to be excluded
In engineering it often comes bounded by
the project itself
18/07/2015
Dr. Felipe Orihuela Espina
39
The research questions
They represent open problems regarding the phenomenon of
interest
Arguably the most important part of your protocol
They ought to guide your research
All goals (main and specific) are collateral consequences of
them
All experiments are driven to answer them
All experimental hypothesis are stated to (educatedly) guess
about them
All conclusions are stated to satisfy them
A rather bad habit is to state them (just because you’ve been
told to), and ignore them the next minute…
18/07/2015
INAOE
40
Goals; Main and specific
The main goal states what is to be
achieved during the thesis
It is a long term goal
It may fall BEYOND the reach and limits of the
project
A thesis may be part of a more ambitious or bigger
project
…that’s why establishing the limits is so critical
It should be stated in one or two paragraphs
at most.
18/07/2015
Dr. Felipe Orihuela Espina
41
Goals; Main and specific
Main goal
Here at INAOE they like it as:
“Develop an algorithm that bla, bla, bla… and that is
competitive with the state of the art”
That is only acceptable if you understand that’s NOT
the real goal but only a collateral goal
The real goal in science is to understand a
phenomenon (that includes computer science!),
…that is; to establish a relationship between dependent and
independent variables under given constraints (i.e. controlled
variables)
…therefore developing/implementing anything is an
engineering goal
18/07/2015
Dr. Felipe Orihuela Espina
42
Goals; Main and specific
Specific goals
Short term goals
They will be covered during the thesis
They may include developing specific
tools/algorithms etc
They often/should also include validation as
one of them
Each one of them has to be describable in at
most 1 paragraph
It is not enough to state them, it is necessary
to describe them
18/07/2015
Dr. Felipe Orihuela Espina
43
Theoretical and reference frameworks
This/These chapter/s summarise/s all foundational
and similar techniques related to the thesis
Theoretical framework: All necessary background to
make the thesis self-contained; transdisciplinary
Reference framework: Literature review on the state
of the art of the phenomenon under study
Beware! Depending on the scientific discipline these
may receive different names and be reported in
different manners; yet they ought to be always
present
For instance, the reference framework in computer science is
often called related work.
18/07/2015
Dr. Felipe Orihuela Espina
44
Theoretical and reference frameworks
In writing these:
Analyze, summarise and criticise the existing
knowledge
Acknowledge the authorship of other authors
(otherwise it is plagiarism)
Ensure that you cite all key works in your area
Often, at the time of writing your protocol
experiments may not have been run1 but a good
knowledge of the topic is expected. This includes:
Domain
Subdomains
1.
At INAOE this is NOT the case; you are expected to have some preliminary results by the
time you write your protocol
18/07/2015
Dr. Felipe Orihuela Espina
45
Methodology; materials and methods
Planning and description of tasks and
experiments
They should refer to the achievement and
solving of the research questions and goals
That does NOT mean 1-to-1 mapping!
It is important both; identification and
description
The more detailed, the easier will be to carry
out them later on
18/07/2015
Dr. Felipe Orihuela Espina
46
Methodology; materials and methods
Scheduling
This is the chronological timing in weeks or
months of the expected times when the tasks
will be carried out.
Always leave enough time for reading and
writing
As a rule of thumb; reading NEVER ends
The Gantt chart or chronogram is often a
good tool to express your scheduling
18/07/2015
Dr. Felipe Orihuela Espina
47
Publications derived from this thesis
If by this time, any publication has been
secured then indicate.
…but remember, no one expects that you
have yet publish a lot…
18/07/2015
INAOE
48
Conclusions
At the time of writing the protocol it is early
to have final or definitive conclusions
…so these do not refer to the intermediate
results in case you have some
… they more often refer to the feasiblity of the
thesis, the foreseen hinders, a brief
discussion over the literature read so far, etc
It should state the impact of your thesis
18/07/2015
Dr. Felipe Orihuela Espina
49
References
Cite ALL relevant references read so far
…so that the thesis committee can assess whether you are
reading sufficiently and effectively
Be coherent with the style format thorughout the
document
Include:
Scientific papers
Books
Other thesis (regardless of the degree)
Technical reports
Web sites
Other sources of information; private communications from other
researchers, public documents, standards, patents, etc
18/07/2015
Dr. Felipe Orihuela Espina
50
PARTS OF A THESIS
18/07/2015
INAOE
51
Parts of a thesis
Title
Abstract
Preliminaries
Table of contents, figure and table listing, abbreviations and acronyms
Chapters
Introductory chapter
Theoretical and reference frameworks
Experimental chapters
Introduction
Methodology
Description of the work developed
Results and analysis
Discussion of results
Conclusions, Recommendations, and future work
References
Appendixes
Others
Index of terms, Source code, maps, circuits, etc
18/07/2015
INAOE
52
Parts of a thesis
The thesis final document is rather different
from the protocol
The protocol describes what is it intended, what
tantatively is it going to be achieved, and a plan
for it. It emphasizes the research questions
In constrast, the thesis describes what has been
done, the results and the conclusions. It
emphasizes the findings and scientific evidence
brought
…consequently, these documents have a
different structure.
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
53
Parts of a thesis
The thesis document, as a scientific
document is far more strict than the
protocol
Normally every institution imposes a
certain structure and format to be followed
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
54
Parts of a thesis
Title page
Chapter numbering and structure is indicative
Abstract
Prelimaries
Ch. 1: Introductory chapter
Ch. 2-3: Theoretical and reference framework
Ch. 4-7: Experimental chapters
Ch. 8: Discussion, Conclusions and future
work
Appendixes
References
Other material
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
55
Title
The title must strictly stick to the
developed research
…and not to the phenomenon or general field
investigated
Accuracy should reign over generality
Be scrupulously precise in the use of
technical terms
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
56
Title
A weak title might be misleading
…specifically, it might lead to false
expectations
…and that include your panel!
A good title might help a correct search
Rules are similar to those of the protocol
…however this is final! This is what would be
indexed by engines
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
57
Title
Do not worry if title is a bit long
…as long as it is justified
…but if you find yourself having a rather long title,
suspect that you are not being concise
If the more important part of the research is
the methodology itself over its application,
then subtitles are often a convinience
Example: Database access through the web
using ASP: Application to Probesi
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
58
Abstract
The abstract opens your thesis document.
In a thesis, it is about 1 pg long
After reading the abstract the reader must know:
What was the problem before this thesis
What has been your contribution
The highlights of your findings
How this thesis impact science
We will see a section about writing abstracts later
on
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
59
Ch.1: Introductory chapter
The first chapter is often an introductory chapter.
It would be expected to cover the following:
Project framing; registry number, funding, conflicts of interests,
Preliminaries and description of the phenomenon
Justification and motivation
Research questions
Goals
Limits
Research hypothesis
Contributions
Publications derived from the thesis
Assessment; how should the thesis be understood and
evaluated
Thesis outline
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
60
Ch.1: Introductory chapter
By the end of the first chapter, the reader
should know exactly:
what have you addressed
what have you achieved
how this work moved science forward
It may be appropriate to use your protocol
as an starting reference
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
61
Ch.1: Introductory chapter
Preliminaries include a brief yet fully
descriptive introduction to the topic,
problem and open questions
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
62
Ch.1: Introductory chapter
Justification and motivation
Indicate the need for your work
It is not the reason why you want to do the
thesis, but the society/market need to gain
that knowledge
For an engineering project, a market analysis is
usually a good starting point, although that
analysis is not part of the project itself
It will be necessary to expand over what you
wrote in your protocol
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
63
Ch.1: Introductory chapter
Research questions:
Hopefully they will be fundamentally the same
that you stated in your protocol
…but this is research! Justified deviations are
acceptable
18/07/2015
INAOE
64
Ch.1: Introductory chapter
Goals and limits
Restate your original goals if necessary but try to
keep them as the originally stated in the protocol
Time and money affecting the goals constraints
should be explicitly declared
In engineering, and as well in computer science,
they should be quantitative if possible
Limits can still be described in terms of what it
has not been done
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
65
Ch.1: Introductory chapter
Goals and limits
Not complying with a goal is a severe fault if not
justified
Again, justified deviations are acceptable
…but even more acceptable is to stick to original statements
and describe why haven’t they been achieved as planned
A time limitation is, more often than not, insufficient
18/07/2015
…that’s why you put your limits in the protocol!
A deviation from the topic is a weak justification.
In that case, a new protocol would have been required
Complications arise during research and errors on the
hypothesis based on new evidence collected during
the thesis are strong justifications
That is what the thesis is all about!
©Dr. Felipe Orihuela Espina (2006)
66
Ch.1: Introductory chapter
Goals and limits
Goals and limits should have been defined a
priori,
…but truth is many do “adjust” them to what
has been achieved
Watch out! This ensures that you comply but it is
often considered dishonest.
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
67
Ch.1: Introductory chapter
Research hypothesis:
There may be more than one.
This is what you defend
Defend also include refuting!
A thesis is the confirmation/refutation of a
hypothesisw
Without yet getting into details (you’ll get to
this later), briefly highlight which hypothesis
are correct or wrong
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
68
Ch.1: Introductory chapter
Contributions
Indicate what new knowledge have you
generated
If collateral achievements are worth (specific
algorithms, databases/corpus, tools, etc), then
state these as well
These by themselves are insufficient to guarantee
a degree
18/07/2015
INAOE
69
Ch.1: Introductory chapter
Publications derived from the thesis
Include all publications derived from the thesis
Clearly separate the journal from then conferences
Clearly separate peer-reviewed ones from others.
Show that you have read adequately by publishing
your theoretical and reference frameworks as a
review
Rule of thumb;
A thesis supported by outstanding record of publications is a
safe shot
A thesis without published back up would be thoroughly
questioned and is unlikely to be enough to grant the degree
18/07/2015
INAOE
70
Ch.1: Introductory chapter
Assessment
Do not underestimate including a good
assessment.
Assessment is not about telling the panel the
qualification you expect for your work
…but about telling the reader how should the
thesis be interpreted and contextualized
The reader should be guided to what you are
interested in him to get from the thesis
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
71
Ch.1: Introductory chapter
Thesis outline
Spend a few lines delineating the structure of
the rest of the document
Example: “Chapter 2 describes …”
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
72
Ch.1: Introductory chapter
Acknowledgements may be part of this introductory chapter, or
made separately in a special section outside
Acknowledgements are expected to include (and in this order)
Yuor supervisor/s – regardless of whether it is deserved
Your sponsor – whoever funded you (research council, business, etc)
Your external (and perhaps even the internals) members of the
committee and panel – for the review!
Other external researchers that may have significantly help you with
your thesis – in whatever aspect (solving doubts, giving ideas, proof
reading, etc)
Other members of your research group who might have been invoved
with your work
Any other person which might have been important for your thesis
Finally, anyone else you want to thank (your family, your
girlfriend/wife/partner, your cat, aunties, etc)
Sometimes, institutions formatting rules indicate the position where
the acknowledgements may appear
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
73
Preliminaries
Preliminaries include:
Table of contents
Figure and table listing,
Abbreviations and acronyms lists
18/07/2015
INAOE
74
Ch. 2-3: Theoretical and reference
frameworks
Theoretical framework
It gives a broad view of all the knowledge related
to your thesis ensuring it is self-contained
This is where your panel will assess your “peripheral
vision”
Includes the knowledge accepted by the research
community and all the theories that will support
our work
It does not yet include any of your work, but some
cirticism on the evidence thrown by your thesis is
acceptable (and encouragable)
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
75
Ch. 2-3: Theoretical and reference
frameworks
Theoretical framework
Do not include things which will not be used
Example: If you include an alternative theory upon
which your thesis is not based, it is because you
will contextualize your finding against it
Do not throw everything you know here;
It is not about showing off or boasting how much
you know
…yet ensure it transpires that you are a true expert
on your field
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
76
Ch. 2-3: Theoretical and reference
frameworks
Reference framework
This is now fully focused on your specific topic
This is where you show the reviewer that you:
Know about your topic everything worth knowing
Understand the subtleties of every aspect of the problem
Analyze, discuss, criticise what other authors have
done
Not everything published is true or correct
Good criticism is constructive
At INAOE a comparative table is often expected, but
this is not the only way to summarise…
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
77
Ch. 2-3: Theoretical and reference
frameworks
Reference framework
Here is where you defend why you favoured
certain techniques or solutions over others.
Defending your choice does not mean refuting
other theories or look down on other’s work
Indicate the strengths and weaknesses on every
work considered
Be open to critics to your own choice
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
78
Ch. 2-3: Theoretical and reference
frameworks
Reference framework
The emphasis of your defence should be
proportional to the relevance for your work
Example: If your work is not about optimization, but
for one of your experiments you needed
optmization; it doesn’t matter too much which
optimization have you used. Just indicate the one
you have chosen and briefly why you consider it to
be the appropriate.
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
79
Ch. 2-3: Theoretical and reference
frameworks
Reference framework
The emphasis of your defence should be
proportional to the relevance for your work
(Cont.)
Example: On the other hand, if you are working in
optimization, highlight why have you favoured that
technique over other alternatives. Clearly describe
the pros and cons in an objective manner (do not
try to bias the reader towards your choice; he may
be a strong defender of the alternative!!)
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
80
Ch. 2-3: Theoretical and reference
frameworks
Reference framework
Show that you are up-to-date!
Ensure thatb you have the latest references on the field
Make extensice and intensive use of research
which has been peer reviewed
Avoid using dubious references (specially from the
internet)
Be comprehensive
A common mistake is to forget about some
author/group’s work; an easy door for attack from the
panel!
Use more than one reference for conflicting concepts
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
81
Ch. 2-3: Theoretical and reference
frameworks
Reference framework
Try to follow some scheme, organization or
flow
Chronological order: from the oldest to the more
recent
By topic: First everything related to one subtopic,
then move to the next subtopic
By acceptance: Review first the most widely
accepted theories and models
By similarity or relevance to your topic; whether by
convergence (most distant first) or by divergence
(most similar first)
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
82
Ch. 4-7: Experimental chapters
Experimental chapters are the
fundamental/main part of your thesis
This is your work!
Each chapter is dedicated to answer
one research question
Concomitantly, each chapter may comply with
some specific goal or afford a specific
contribution
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
83
Ch. 4-7: Experimental chapters
The emphasis of each chapter will however
depend on your thesis aim:
Emphasize methodology if novel or scarcely used
Emphasize experimental design and rigour if
exceptional
Remember! Experiments have to be carefully designed
and planned even if this is not the goal of the chapter
Emphasize the research hypothesis associated to
the question
Emphasize the contribution of the chapter
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
84
Ch. 4-7: Experimental chapters
The emphasis of each chapter will
however depend on your thesis aim
(Cont.):
Emphasize differences in experimental design
against other authors’ choices
Emphasize optimization of resources if
applicable
Sometimes a better experiment is easy to design
or foresee but practical limitations prevent their
implementation
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
85
Ch. 4-7: Experimental chapters
The emphasis of each chapter will
however depend on your thesis aim
(Cont.):
Emphasize the description of the algorithms
and their parameterization if the experiments
involve a collection of simulations
Emphasize the sequence of the experiments
and simulations if thie is relevant
Do not assume that commutative does apply
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
86
Ch. 4-7: Experimental chapters
The description of your experiments ought
to be:
scrupulous,
rigorous,
exhaustive
utterly/brutally honest
guaranteeing replicability of your results
Non reproducible results are simply invalid!
This is what science is all about!
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
87
Ch. 4-7: Experimental chapters
The description of your experiments ought
to be (Cont.):
Indicate not only the factors you have
manipulated (independent variables) but also
everything that is kept constant (controlled
variables)
Painstakingly describe all pre-processing,
processing and analysis of your data, and this
includes the statistics for hypothesis testing
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
88
Ch. 4-7: Experimental chapters
There is a full theory about experiment
design…
It ensures your experiment reduces bias, cost
and time
It has a strong statistical connotation that at
the end of the day is an optimization problem;
aiming to solve your unknowns with minimal
“effort” measured in some way
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
89
Ch. 4-7: Experimental chapters
Minutely describe all results with clear distinction
between quantitative and qualitate statements
Do NOT include any judgement or adjective
Stay objective at all times
Stick to your results and do not extrapolate
This is just a reporting exercise, the time to speculate will
come in your conclusions chapter if you want
Do not be afraid of negative findings/results
Most times, science advances through negative findings
…although of course these requires even more thorough
attention to the way they are reported.
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
90
Ch. 4-7: Experimental chapters
Use but don’t abuse stats
Example: If you did 10000 simulations you do
not need to report all 10000 results
independently, just use descriptive stats
Graphs are a good way to summarise info
Tables are more explicit that graphs and plots
but often more difficult to read.
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
91
Ch. 4-7: Experimental chapters
When using graphs/plots:
They must be representative of the
information you want to show
The must be descriptive of the knowledge you
want to convey
They must be clear, readable, properly
labelled, quickly interpretable and selfcontained (independent of the text)
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
92
Ch. 4-7: Experimental chapters
When using graphs/plots:
Choosing the right plot might not be trivial
…indeed! There is a whole research dedicated to
data visualization techniques
Each plot type is better to highlight some kind of
information communication
Data visualization is critical
Sometimes visualization is improved if data are
previously transformed (e.g. use of logarithmic
scale)
Sometimes it may be worthy to ghost some data to
emphasize some other aspect of the information
(e.g. outlier removal or hidding, axes cropping)
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
93
Ch. 4-7: Experimental chapters
When using graphs/plots:
If you have more than 3 dimensions/variables
you may want to consider dimensionality
reduction techniques (e.g. PCA)
3D plot interpretation maybe impressive but
not necessarily easy; at the end of the day,
paper is bidimensional
Never, ever, sacrifice clarity for the sake of a more
aesthetically beautiful plot
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
94
Ch. 4-7: Experimental chapters
If the experiment is an indirect measure of
a non-observable phenomenon, then
clearly state your forward/inverse model
relating observable information to nononservable information
Example: Image reconstruction
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
95
Ch. 4-7: Experimental chapters
NEVER, ever, manipulate your results
As serious offence
Do not force regressions, nor distributions
Respect techniques assumptions and confine
your interpretation to these
Victimism is not an option in science:
Example:
You say: “I cannot have more data. In my field collecting
18/07/2015
data is time consuming/costly/etc”
What I hear: “I’m lazy. I couldn’t care for carefully designing
my experiment to fit my circumstances”
You say: “Should I have more data, my results would have
been significat/better/more solid”
What I hear: “I have no idea about stats, so instead of doing
the right thing i.e. learning stats, let’s do the easy thing,
blame what everyone else does…”
©Dr. Felipe Orihuela Espina (2006)
96
Ch. 4-7: Experimental chapters
In engineering, you ought to include:
Error analysis
Cost analysis
Many theoretically valid solutions may not be
feasible for practical constraints, and cost is
certainly one of these
Pay special attention to:
Units
Type of variables
Not all variables admit the same operations
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
97
Ch. 4-7: Experimental chapters
In engineering it is also necessary to
express intermediate needs and structures
Example: Civil engineering; in improving a
motorway, you should not stop the motorway
for three months. You have to provide a
temporal diversion.
Example: Computer science: If you are
updating the structure of a database provide
an intermediate repository and ensure you
have a plan for restoring/adding this
information to the new database.
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
98
Ch. 4-7: Experimental chapters
Discuss your results
Subjectvity is not welcome, especially in exact
sciences
Be cautious about leap of thinkings
3 samples aren’t enough to demonstrate a trend
Highlight both strengths but also weaknesses
Compare, compare, compare
Comparing is the basis of validation
Important: Include those approaches who are
“better” than you. Not doing it opens a door for
attack from your panel.
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
99
Ch. 4-7: Experimental chapters
Analysis is more than just stating the obvious;
it is demonstrating that you understand the
subtleties and intricacies of your results
Example: You have a data series following a
Gaussian distribution
Does it fit to what you expected? Maybe you were
expecting some other distribution
Is standard deviation too large? What about standard
error; too small?
Is your positive result of the test a false positive?
Is sample size adequate? Not too small, not too large
Is this relation likely to be spurious?
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
100
Ch. 4-7: Experimental chapters
Demonstrate that you know how to interpret your
results
Example: ¿Are your data against the laws of physics?
¿Turtles running 110m obstacles in 3 sg?
¿2+2=38?
¿Why did my genetic algorithm took 15 days to run if only
had a population of 3, 5 generations and fitness function was
trivial to evaluate?
Did you design your experiment correctly? Was it
18/07/2015
actually executed correctly?
Since it is almost impossible to run a “perfect”
(unbiased) experiment, what biases might be
responsible for these observations?
©Dr. Felipe Orihuela Espina (2006)
101
Ch. 4-7: Experimental chapters
Do not anticipate conclusions!
Consider alternative hypothesis
If negative results, indicate why you think it
failed and how would you tackle the issue
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
102
Ch. 4-7: Experimental chapters
Particularly for software
Just because it run once, it does not mean it would
run always
Different hardware rounds in different manner even under the
same OS,
If (pseudo-)random elements are present, initial conditions
may differ
Home/lab testing is never exhaustive
I don’t care if you took 8 months without eating or sleeping
A good bunch of empirical simulations is not an analytical
proof of model checking!
Blind/wild trial and testing? Or thoroughly/carefully
18/07/2015
plan tests?
Did you allocate sufficient time for design and
analysis (commesurate with the application size)?
©Dr. Felipe Orihuela Espina (2006)
103
Ch. 4-7: Experimental chapters
In science, you can’t hide from stats!
…well, maybe you can: “If you need stats to
prove your hypothesis, you ought to have
done a better experiment” Lord Ernest
Rutherford
Ensure you have the correct statistics, both
descriptive and inferential, according to your
experimental design
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
104
Ch. 4-7: Experimental chapters
Pay special attention to results changes
under different conditions
Do you get the same or congruent results?
Are they disparate?
Is what you have measured what you really
wanted to measure? Just a proxy?
Check your boundary conditions
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
105
Ch. 4-7: Experimental chapters
A few unknowns can be estimated in
different ways with experiments having
slightly different (complementary) point of
view
This is often a good way to confirm a result
Just ensure your results do back up each
other
If you got different results, then question why?
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
106
Ch. 4-7: Experimental chapters
When comparing to other works
Make fair comparisons; homogeneous
groups, comparable treatments, etc
In case of doubt, always give advantage to
the other methods
Sometimes, repeating someone else’s
experiment is a good way to check your
approach (…as well as an excellent
opportunity to verify the other author’s results)
If your results are outstanding, then be
skeptic!
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
107
Ch. 4-7: Experimental chapters
Discussion: This is where you evaluate
your findings and contextualize them with
accepted knowledge
Identify your own weaknesses, assumptions
(especially those hidden or not obvious),
biases, etc
Be critical and objective to your own work
Contextualize with accepted knowledge
(nomological validity)
A strong discussion saves a weak thesis; a weak
discussion spoils a good thesis
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
108
Conclusions, Recommendations, and future
work
Conclusiones: This is a summary of your
findings
Think how have you contributed to science
It is not a mere summary of your results, but a
thought on their implications
A common way to report them is by research
question (section) and then by specific finding
(subsection)
It is here where you state if you have confirm
or refute your original hypothesis
i.e. where your hypothesis becomes a thesis
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
109
Conclusions, Recommendations, and future
work
Recommendations:
Highlight some of the difficulties and
challenges you faced during your
experimentation and how did you solve them
Look beyond your own work; think how it does
impact other areas of science and technology
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
110
Conclusions, Recommendations, and future
work
Future work: A good research always led
to new open questions
Show that you know how to continue your
research (when you no longer have a
supervisor)
Identify ways in which your work can be
modified, improved, enlarged, generalized,
particularized, further verified/validated, etc
A good way to extend your work is simply by
relaxing/attending some of the assumptions
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
111
Conclusions, Recommendations, and future
work
Future work:
Identifiy how your work can seed other works
Identify gaps that have to be more thoroughly
studied
Launch new hypothesis on those yet
unresolved issues
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
112
Appendixes
It is also where you append your thesis with
supplementary material necessary for the
thesis to become self-contained
Long mathematical proofs
Specification sheets,
Very large tables
Consent forms
Maps
Budgets (compulsory for engineering projects)
Code listing
Photographs
etc
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
113
References
Regardless of the chosen style (perhaps
imposed by the formatting rules of the
institution), it is necessary
Sort your references in some way
Keep a coherent citing/referencing style all
throughout the thesis
Rule of thumb; In a 4 year thesis you shall
read about 450-600 papers of which 1/3-1/2
will be useful and will be included in your
thesis
…less than that, and you haven’t read enough!
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
114
References
Common sorting.
By order of appearance in the text (very
common in papers)
By surname of the first author
By year
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
115
References
Style
There exists some standard styles: Chicago,
Vancouver, Harvard, etc
In computer science the Vancouver style is
perhaps the most widely used
These dictates not only the reference format,
but also the citation format
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
116
Throughout the thesis
Remember!:
Avoid PLAGIARISM. The most severe offence in
academic work
Use citations and quotations appropriately
The rule is simple; Original authorship of every piece of work
must be crystal clear
Avoid rambling and repeating concepts
Use references to equations, sections, figures, tables, etc
The thesis is written for your peers
Avoid including trivial or basic material
…but also remember that your peer may not be an expert on
you specific field.
18/07/2015
©Dr. Felipe Orihuela Espina (2006)
117
WRITING ABSTRACTS
18/07/2015
INAOE
118
Writing Abstracts
There exist 3 major types of abstracts:
Descriptive
Informative
Critical
A fourth type is the so called extended
abstracts; a longer than normal informative
abstract that acts as a substitute of a full
document
18/07/2015
Dr. Felipe Orihuela Espina
119
Descriptive Abstracts
Short – Usually less than 100 words
Include:
Purpose of the work (Goals)
Methods
Scope
Does not include:
Results, conclusions nor recommendations
The reader will have to read the full
document to decide whether it is relevant for
him/her
The kind of abstract you include in a review
18/07/2015
Dr. Felipe Orihuela Espina
120
Informative Abstracts
Mid size - Between 200 words and 1 pg
Include:
Background
Aims
Methods
Results
Conclusions and impact
Communicates the content of the document.
Sufficient for a reader to decide its relevance.
The kind of abstract you include in a paper
18/07/2015
Dr. Felipe Orihuela Espina
121
Critical Abstracts
Long – 500 words to 1000 words
A non structured short summary of the
document
Highlights the take away message of the
content
The kind of abstract you include in your
thesis
18/07/2015
Dr. Peter Hancox/Trad. Dr. Felipe Orihuela
Espina
122
Characteristics of a good abstract
Concise
Communication is effective and efficient
Informative
Regardless of the type the reader gets a quick
overview of the contents of the document
Connected
Flow is smooth and without breaks
Conservative
Does not include new material over the material
contained in the main document
18/07/2015
Dr. Felipe Orihuela Espina
123
Common errors
Title repetition
Well, the title is a different part of the
document
Boring
Wasted invaluable space
Starting with “This work…”
Isn’t it obvious that you’ll talk about this
document?
18/07/2015
Dr. Peter Hancox/Trad. Dr. Felipe Orihuela
Espina
124
Common errors
Making external references
Title and abstracts should conform an selfcontained unit
They have to be used for indexing the
document
Avoid references to literature and figures or
tables
18/07/2015
Dr. Peter Hancox/Trad. Dr. Felipe Orihuela
Espina
125
Common errors
Use of acronyms and abbreviations
Leave those for the main text
Only extremely well known acronyms may be
acceptable; USA, UK, etc
Do you know what RTE, ATR, GSM, WAM or
LFG mean? Well, they are extremely well
known in their areas…so why should you
expect that someone knows the
acronyms/abbreviations in your field?
18/07/2015
Dr. Peter Hancox/Trad. Dr. Felipe Orihuela
Espina
126
THANKS, QUESTIONS?
18/07/2015
INAOE
127