1. Scientific Posters
Principles of Design
Tsai-Shiou Hsieh
Macaulay Honors College @ Queens College
City University of New York
tsaishiou.hsieh@qc.cuny.edu
(With some slides shared by current and former ITFs:
Russell Hogg, Michael Porter & Craig Willse. Thank you all!)

2. What is a Scientific Poster?
Visual means for communicating research to an
academic or professional community.
A summary of research that serves to create
interest by highlighting most important
findings.

3. Requirements
 Each group must produce one poster.
Posters can be made with Keynote or
PowerPoint.
At least one side of the poster should be no
longer than 36”. (Recommended dimensions: 48” x
36” or vice versa).

4. Organizing Contents
Title / Authors/ Affiliations
 Abstract
 Introduction / Background (Personas)
 Methods
 Results & Discussions (Challenges)
 Conclusions (Solutions)
 References
 Acknowledgement

5. Some Basic Graphic Guidelines
Images should guide the overall layout, not the text.
Avoid cluttering the poster (graphs, photos, etc.)
Watch your color contrasts.
Make sure all components are aligned properly.
Use some kind of underlying structure!

6. Now….. Be prepare to see a
BAD example of poster….
Ready?

7. We describe the philosophy andWe describe the philosophy and
design of the control flowdesign of the control flow
machine, and present the resultsmachine, and present the results
of detailed simulations of theof detailed simulations of the
performance of a singleperformance of a single
processing element. Each factorprocessing element. Each factor
is compared with the measuredis compared with the measured
performance of an advanced vonperformance of an advanced von
Neumann computer runningNeumann computer running
equivalent code. It is shown thatequivalent code. It is shown that
the control flow processorthe control flow processor
compares favorably in thecompares favorably in the
program.program.
We present a denotationalWe present a denotational
semantics for a logic programsemantics for a logic program
to construct a control flow forto construct a control flow for
the logic program. The controlthe logic program. The control
flow is defined as an algebraicflow is defined as an algebraic
manipulator of idempotentmanipulator of idempotent
substitutions and it virtuallysubstitutions and it virtually
reflects the resolutionreflects the resolution
deductions. We also present adeductions. We also present a
bottom-up compilation ofbottom-up compilation of
medium grain clusters from amedium grain clusters from a
fine grain control flow graph.fine grain control flow graph.
We compare the basic blockWe compare the basic block
and the dependence setsand the dependence sets
algorithms that partition controlalgorithms that partition control
flow graphs into clusters.flow graphs into clusters.
Our compiling strategy is toOur compiling strategy is to
exploit coarse-grain parallelismexploit coarse-grain parallelism
at function application level:at function application level:
and the function applicationand the function application
level parallelism islevel parallelism is
implemented by fork-joinimplemented by fork-join
mechanism. The compilermechanism. The compiler
translates source programstranslates source programs
into control flow graphs basedinto control flow graphs based
on analyzing flow of control,on analyzing flow of control,
and then serializes instructionsand then serializes instructions
within graphs according to flowwithin graphs according to flow
arcs such that functionarcs such that function
applications, which have noapplications, which have no
control dependency, arecontrol dependency, are
executed in parallel.executed in parallel.
We apply a parallel simulationWe apply a parallel simulation
scheme to a real problem: thescheme to a real problem: the
simulation of a control flowsimulation of a control flow
architecture, and we comparearchitecture, and we compare
the performance of thisthe performance of this
simulator with that of asimulator with that of a
sequential one. Moreover, wesequential one. Moreover, we
investigate the effect ofinvestigate the effect of
modeling the application onmodeling the application on
the performance of thethe performance of the
simulator. Our study indicatessimulator. Our study indicates
that parallel simulation canthat parallel simulation can
reduce the execution timereduce the execution time
significantly if appropriatesignificantly if appropriate
modeling is used.modeling is used.
We have demonstrated that toWe have demonstrated that to
achieve the best executionachieve the best execution
time for a control flowtime for a control flow
program, the number of nodesprogram, the number of nodes
within the system and the typewithin the system and the type
of mapping scheme used areof mapping scheme used are
particularly important. Inparticularly important. In
addition, we observe that aaddition, we observe that a
large number of subsystemlarge number of subsystem
nodes allows more actors tonodes allows more actors to
be fired concurrently, but thebe fired concurrently, but the
communication overhead incommunication overhead in
passing control tokens to theirpassing control tokens to their
destination nodes causes thedestination nodes causes the
overall execution time tooverall execution time to
increase substantially.increase substantially.

8. Photo Sources & Tips
 Take pictures yourself when possible!
 Flickr (www.flickr.com)
e.g.: http://www.flickr.com/photos/dnwallace/2635909272/
 Fotopedia (www.fotopedia.com)
 Attribution / Creative Commons
 Don’t enlarge photos

9. Text Guidelines
Break text up with bullets or numbers
Indenting shows subordination
- As in this example
Avoid lengthy paragraphs talking about why you did what
you did and whether you dislike positivism because there
is such a thing as reality out there and it operates in a
certain way and we should be able to access that in some
shape, form, or fashion and besides it’s all from some
stuffy old dead guy thinking too hard, anyway.

10. Text Guidelines
Be sure your letters stand out against the background.
Dark gray is a good color, but not against black background.
Keep your color palate simple
Rainbows are good for Halloweens, not for conferences
Use fonts people can read- both types and sizes!
This is very artistic for making a card, but not really good for making posters
• Titles, subtitles and headings: 40 to 70 point font
• body text: at least 24 point font
• TRIM EVERYTHING THAT’S NOT DIRECTLY
PERTINENT

11. Text: Less is More
Teeth are ideal for
studying life history
because they grow
incrementally, are not
remodeled during an
individual’s lifetime, and
are not highly subject to
environmental stresses.
Teeth & Life
History
 Incremental growth
 Not remodeled
 Resistant to
environmental
stress

12. Left to Right, Top to Bottom Flow
Title & Authors

13. Left to Right Flow in Rows
Title & Authors
Part 1
Part 2
Part 3

14. Centered Image & Peripheral Explanations
Title & Authors

15. Centered Explanation, Peripheral Images
Title & Authors

16. Final Words
 Work on your research first!
 Rich contents. Simple design
 Visuals are good, but include only
meaningful, theme-relevant tables, charts
and photos (don’t use them as space-filler).
 Draft early. Multiple revisions!