This folder includes the files to recreate most of the results from

Meir, von Dassow, Munro, and Odell. Robustness, flexibility, and the
role of lateral inhibition in the neurogenic network. Current Biology, 2002.

We wrote and used the files with version 0.8 of Ingeneue, and they should
work with the version currently available for download from www.ingeneue.org.

These files are all copyright 2001 by Meir, von Dassow, Munro, and Odell. Permission
is granted to use them for any non-commercial purpose, though please site the above
paper in any publications resulting therefrom.

Below are brief instructions for how to use the files to generate the results in the
paper.


The files:

Three directories contain the files for the three different patterns we
discuss in the paper. The files directly inside each directory are the "network"
files which define each of the networks. A fourth directory contains
contains the "iterator" files we used to explore that pattern.

Random searches:

To randomly search for parameters that make a particular pattern, run the
following sequence:

Load a network file for that pattern (either 2 cell or 7 cell - we did not create the
file to randomly search for line patterns, though we could tell you how to do it). If
you are working with the seven cell pattern, load the file without "Both" in it.

Load the neuro1stFilter.iter file from the iterators folder corresponding to the
pattern you loaded.

Run that iterator. You'll see the number of parameter sets the program has
tried printed in the console window. Each time it finds a successful parameter set,
it will print a note about the success.

When you've seen enough parameter sets go by, quit the program.

Rerun Ingeneue, load the network file again. If you are working with the 7 cell
pattern, load the network file WITH "Both" in it. Find the output file from the first
run (which should be in a folder called output, inside the folder containing the
Ingeneue program).

Load the neuro2ndFilter.iter file. Run that. This filters all the initial successes
using a better integration algorithm.

Quit the program. Rerun Ingeneue. Load the network file (with "Both" if seven cell).

Load the new output file, produced by the second run above.

Load the neuro3rdFilter.iter file. Run that. This filters all the previous successes
through a 1000 minute test (earlier tests were for 300 minutes).

This final output file is the one you want. At the bottom you'll
see a message about the number of successful parameter sets you found. If you look in
the original output file (from the first run above), you'll see a "count" of the number
of parameter sets the program tried. The successes divided by the count gives our
solution frequency.

Mating:

To mate parameter sets together, load a model as above.

Load an output file from above that has a bunch of successful parameters in it.

Load the matingtrial.iter file for the appropriate pattern.

Run that iterator. It will do 10,000 matings between parameter sets randomly
chosen from the output file you loaded. At the end, you can read the mating success
rate from the console window, or from the bottom of the output file that
the mating algorithm makes.

Mutational expansion:

To expand around a parameter set using our mutational expansion algorithm, open
the network file in a text editor.

Open the output file containing the parameter set you want to expand around.

Copy the parameter set from the output file into the network file, replacing the list
of parameters currently in that network file.

Save the network file with the new parameter values.

Run Ingeneue and load the network file.

Load the mutateExpand.iter file for that pattern. Run it. This can take a 
very long time to finish, depending how good a solution you started with 
and how you set the parameters in mutateExpand.iter.

When it finishes, quit Ingeneue. Rerun Ingeneue, load the network file, and load
the output file from the mutational expansion.

From the wheel diagram window, select Summary from the Statistics menu.

The summary file that it outputs will contain the ranges of each of the parameters
from within the cloud of parameter sets generated by the expansion. It also contains
a measure of the largest euclidean distance between any two parameter sets in the
cloud.

Flexibility of parameters:

To try parameter sets from one pattern with another pattern, load the network file
for the new pattern. Load the output file generated by random search (or some other means)
from the original pattern. Load the neuro2ndFilter.iter iterator file for the new
pattern. Run the iterator.

Parameter transects:

To make transects along each parameter axis, load the network file. 

Load an output file containing the parameter sets you want to make transects around.

Load the Transect.iter file.

Run the iterator. This can take a really long time, and will obviously take longer
the more parameter sets you have in the output file you load.

The output file is pretty dense. The top line for each parameter shows the values it
was given along the transect. The second line for each parameter shows the score the
network received at that value of the parameter. We have another program  that we are 
happy to share which will summarize these files as in the figure in the paper.