"surplus slack" and citing

[Maria C]

Hello Michael!! I am running a search using General Multi-Criteria
Analysis, I am using this objective because I need area and deficit
constrain. I was wandering what is the "surplus slack" and how can
this improve the search??

Is there a place where I can download Ciarleglio, M. (2008). Modular
Abstract Self-Learning Tabu Search (MASTS): Metaheuristic Search
Theory and Practice [dissertation]. University of Texas at Austin,
Texas. Or can you send me the pages were you talk about improving the
search in the General Multi-Criteria Analysis??

The last question is if there is a paper I can use for citing ConsNet,
or how do I cite it??

Thanks!!

Maria.

[Michael Ciarleglio]

Hi Maria,

Thank you for you interest in ConsNet. I'm excited that you are
looking at some of the more advanced features. It sounds like you are
trying to maximize representation given some budgetary constraints
(such as an upper limit on the total area). The current version of
the software is not perfectly tuned for this type of problem (we
focused first on the minimum area problem), but you can use the
"general multi-criteria analysis" tool to help with your analysis. We
use this tool in our research to test different ways of solving
various problems. There are no guarantees that it can solve the
problem perfectly, but with experimentation you may find certain
techniques that work. With more research, we will find better methods
for solving the "maximum representation problem", and we can include
these in later releases of the software.

Since the instruction manual does not go into detail, I'm going to
elaborate on the topics you mentioned.

>>Citations

Ciarleglio, M., Barnes, J. W., and Sarkar, S. ConsNet: New Software
for the Selection of Conservation Area Networks with Spatial and Multi-
Criteria Analyses. Ecography, in press.

Ciarleglio, M., Barnes, J. W., and Sarkar, S. ConsNet - A Tabu Search
Approach to the Spatially Coherent Conservation Area Network Design
Problem. Journal of Heuristics, in press.

>>Dissertation
I'm going to check with the University to see if I can post it here on
Google Groups. Otherwise, I can distribute it individually.

>> More About Surplus Slack:
Surplus slack is not a criteria that can be used to judge the quality
of the conservation area network. It is a marginal measure of how
many representation targets have been exceeded. In the real world,
planners do not need to know about surplus slack, and should ignore it
completely. However, the search engine in ConsNet can use surplus
slack to make better decisions about which solutions to choose next.
Using the surplus slack in this context can greatly improve the
performance of the search, as long as you are solving the "minimum
area problem".

We use surplus slack to help drive the search in the "multi-criteria
minimum area" objective, and it makes the difference between night and
day (in terms of search performance). However, the surplus slack is
not considered when we rank and present the best solutions. The
scores assigned to each solution only include the preferences defined
by planners.

I don't think that the surplus slack will be useful in solving the
"maximum representation problem", such as the one you have described.
It only makes sense to consider surplus slack in the "minimum area
problem" (although you are welcome to experiment). I believe that
with research, we can find another quantity that is similar in purpose
which will lead to substantial gains in the "maximum representation
problem".

>>General Multi-Criteria Analysis:
There will be two parts to a "general multi-criteria analysis"
objective. First, you define the multi-criteria analysis (MCA), which
is basically the weights that are assigned to each criteria. The
primary question is, "Which criteria should I include, and how
important is each one?". Next, you define a constraint gate, which
places restrictions on the types of solutions that you are willing to
consider. As mentioned in the ConsNet manual, it is best to start
with one constraint (such as an upper limit on area). Adding
additional cost constraints (such as an upper limit on the price of
land, or the total human population in the network) should be
straightforward as long as the constraints are "less than". Adding
other types of constraints could lead to confusing behavior (but might
also increase the effectiveness of the search).

>>Tips for the Maximum Representation Problem
We have looked at problems similar to the one you mentioned. I will
describe some of the different strategies we have used. I think it
will require some research to find a really robust solution, but we
have had significant success using the "general multi-criteria
analysis". For the MCA, given a specific target (or choosing from
multiple targets), you might try a combination of the following:

1) minimize the largest deficit (favors the most abundant species)
2) maximize the number of satisfied targets (favors rare species)
3) minimize the total deficit (favors the richest cells)
4) minimizing the shape
5) minimizing the number of clusters

Note that for items 1-3, you can use *any* combination of the
available targets. If you have specific subsets of species that need
special attention, try creating a special target just for them (you
will have to reload the problem with your new target file). For
example, you can create a special representation target just for the
endangered species (zero for the other species), and put more weight
on meeting those targets.

We have found that items 4 and 5 complement each other well. I would
recommend using both.

Some auxiliary goals, which you might want to include in your MCA:
6) maximize total representation (this includes representation above
the specified target)
7) minimize area

Number 7 is kind of a strange recommendation, because minimizing the
area is not your stated goal. But sometimes the search is better off
choosing a solution that has less area, so that it can find more
efficient ways to increase surrogate representation. This may be a
trick to improve the search performance (somewhat like surplus
slack). Finding these tricks, and making them work transparently is
still an open research question.

If it is absolutely required that some (or all) species meet a minimum
representation target, then you can implement this using a
constraint. For instance, if your conservation area network must
include 5% of all species, you should add the following constraint:

-->total deficit - 5% of total surr "less than threshold" 0

Using "total deficit" works better than number of satisfied targets,
due to implementation details. I would also point out that you can
use any custom target here. The following details are very important:
a) the area constraint must come first
b) If you specified a maximum area, then it must be possible to
include 5% of all species within this area. Otherwise, you will never
be able to meet both constraints simultaneously (and the search will
fail). To be safe, the land required to meet the minimum
representation constraint should be substantially smaller than the
maximum area that you specify. Otherwise, the search enough
flexibility to explore different spatial configurations (which may
require more area than you expected). When performing this type of
search, it may be helpful to find a high quality solution that meets
the minimum representation constraint, and then start from that
solution.

*****
OKAY! That was a lot of information. Maria, if you have any more
questions, I can answer them here on Google Groups or through email.

Michael
michael.ciarleglio.googlepages.com
michael.c...@gmail.com