|
Viz Development Team
Dr Steven
Halim
Dr
Roland Yap Hock Chuan
Dr
Lau Hoong Chuin
Felix Halim
Feedbacks
Please drop us your
feedbacks via this online feedback form.
Latest Updates
30 May 2011:
Viz v3 (the final one used in Steven's PhD thesis) is
released. Click here
to download it! No registration is needed
5 Sep 2008: See our
results
page to see what our 'state-of-the-art' LABS (Low Autocorrelation
Binary Sequence) solver: TSv7 can do. These results will be
presented in CP
2008.
|
 |
 |
|
 |
 What
is COP?
What is SLS?
Was SLS behavior mysterious?
Explain basic ideas! |
On the left, Viz version 2 shows
Ro-TS-I versus
Ro-TS-B
on
QAP type B instances.
More details are
in our
results
&
documentations
pages.
Download
and try Viz now!
It is free :),
but you need to
register to
obtain your key. |
Overview:
Now SLS behavior on COP Fitness
Landscape is no
longer too mysterious!
Viz (version history) is an off-line, user friendly,
GUI-based,
Stochastic
Local Search (SLS) engineering suite.
Viz is used in Steven's 6-years PhD Thesis (2 Aug 2004-25 Aug 2010) (click
here to view).
Viz can be used to visually analyze
(white-box) Stochastic Local Search (SLS) (a.k.a.
metaheuristic) algorithms while they are traversing the fitness
landscapes of
NP-hard
Combinatorial
Optimization Problems (COPs). This visualization is problem independent,
which means that it can virtually be used to visually analyze almost any COPs that is attacked by an SLS algorithm. The visualization visualize both the
fitness landscape (FL) of a COP instance and
the search trajectory (ST) of a heuristic and stochastic SLS algorithm on that fitness landscape (the basic ideas behind this
FLST visualization are explained here).
Understanding search trajectory behavior of our SLS algorithm on the COP
instances being attacked empowers the user (algorithm designer) to
design better performing algorithm and focus the parameter space to a
much smaller one. Once the SLS algorithm has been
(carefully) designed and parameter space is (significantly) focused, Viz
can then be instructed to perform full-factorial design (black-box)
on the focused configuration space for even better performing algorithm.
We know that to engineer good
performing SLS, one needs to design, implement, tune,
analyze the SLS algorithm. The visualization in Viz is
as a good tool for the design and analysis parts. The
full-factorial design capability in Viz is a good tool for the tuning
part. This combination is referred as the Integrated White+Black Box
Approach.
Mirror page (which is
outdated, this page is the more recent one):
http://sls.visualization.googlepages.com
|
