Wednesday the 8th of October 2014
*Adapted from the slides of Mark van Rossum, Andrew Stewart and Vincent Valton
available at the IRR wiki:
Taught component (80 credits)
● lectures, tutorials, coursework, exams
● learn established techniques that work
Research component (100 credits)
● do something that’s never been done before
○ study a new problem, develop a new method, etc.
○ probably the most exciting (and hardest) part of MSc
● culminates in you writing a dissertation (~50 pages)
● two courses prepare you:
○ IRR: literature review in area of interest
○ IRP: write a detailed plan for your MSc project
Your review has to be:
● 6 full pages (about 3000 words)
● How many papers will I need?
■ (3-4) in depth,
■ 30 read/understand,
■ 50 skimmed.
● Submit by 4pm, 15th January 2015, no extensions
Pass/Fail is based on:
● Appropriate coverage: did you hit all the important papers in the area?
● Understanding of sources: are you just parroting back what you read?
● Critical evaluation and comparison: beyond “A did X, B did Y”?
● Clarity of expression and presentation: can your friends understand it?
● Attendance of tutorial group meetings: discuss all absences with me
IRR is not:
1. to write about a topic without solid references
2. to simply summarise some papers you have read
3. original research (but instead a novel appraisal of existing research)
Best way to FAIL:
2. do not come to tutorials
*to be confirmed!*
1. (8/10) Introduction + Look for a topic
2. (15/10) Choose topic and select at least 10 papers from it
3. (22/10) Read through at least 5 papers
4. (29/10) Bullet point draft
5. (5/11) Expand some sections of the draft + main reference list
6. (12/11) First complete draft of the review
7. (19/11) Feedback on the review
4pm, 15th January 2015, no extensions
Homework for next week
First goal: identify which area to review
● This can be an iterative process, so better start early!
● Topics can be similar, but they should be unique
1. Seed papers are a good way to start
2. Look online for other topics you are interested in
a. Google Scholar
Choosing a topic
The topic of your review should be:
● Not too general,
○ e.g. “review of Artificial Intelligence”
... 800,000 relevant papers
● Not too specific
○ e.g. “Applications of pancake recipes in robotic automation”
... not a well established field, 2-3 relevant papers
Choosing a topic
You should find at least one recent and widely cited paper.
It should be:
a. published in the last ~3 years
b. at least ~30 citations…
If you have difficulties finding such a paper, the topic might be too specific or it
might have died out.
Choosing a topic
Next week we will discuss your topic and I will help you to generalize it or
narrow it down, if necessary
For next week, find a possible topic and have read at least one strong (recent
and widely cited) paper on that topic.
● By Tuesday the 14th at 12:00, send me an email with:
○ Short description of the topic you have identified (3-5 sentences)
○ At least one reference to the most relevant paper you have found
■ Full name of all the authors
■ Name of the paper
■ Name of the conference/workshop/journal it was part of
■ Year of publication
■ Is it one of the advertised topics? If yes, which one?
You can send me the description and references of up to 3 topics
My email address: P.Pareti@sms.ed.ac.uk
Seminars are not compulsory, so why attending them?
1. Free food (not always though)
2. Great talks from great people
a. Leaders both from academia and the industry
b. You can actually ask questions and maybe chat with them
3. Serendipity. Seminars are great sources of inspiration. If you focus only on
your field, you will miss out on a lot of ideas which would be very useful to
Accessing and Evaluating papers
VPN (access papers from home)
● CORE ranking (mostly for conferences)
● Google Scholar
● LaTex + BibTex
○ You are not forced to used them...
○ … but if you don’t, you are most likely to lose a lot of time
○ IRR is a great opportunity to learn it
Other useful stuff
● Friendly interface on top of LaTex
● Friendly bibliography reference manager