Presently, measurements are made and statistics derived during simulated development. A case can be made for measurements made with a separate program.
Instead of trying to capitalize on the ease of access to information within the program that simulates network development, measurements and the generation of statistics can be done  separately, an approach that resembles the statistical analysis of biological experimental neuronal networks. The results of simulated development (or of simulated activity in a network) can be stored at desired stages. A separate program can use the same file reading functions to retrieve those results and apply measurements. The measurements would closely resemble measurements done in biological networks, e.g. the coordinates of fiber segments may be measured in order to derive angles.

This approach may have some benefits:
It is possible to delegate the task of developing measurements to someone else.
Measurements can be done on a separate computer.
A-posteriori measurements may be done that were not initially conceived of.

In some cases, it may appear that statistics are much more easily obtained during development than a-posteriori, sucn as in cases where recognition of the objects to measure is non-trivial. For instance, if we wish to compare statistics for apical dendrites with those of other dendrites then we need to know which dendrites are apical dendrites. Of course, in order to make measurements within the program that simulates development we need to know in advance which measurements might be of interest. If we know that then we could just as easily include the necessary markers in the network storage format to make recognition trivial for a separate measurement program.

Created by randalk 

Last modified 2006-02-04 05:59 AM