[ USCBP || FARS || up || down || prev || next ][ model || bio || impl || sim || doc ]

3. Network Files I/O


In this section, we will describe various components in the implementation of FARS. The core program, FARS, is implemented in NSL, but it is not a stand-alone program. FARS requires some other programs to generate part of its input files. We will look at these programs, the input/output files of FARS, and their inter-relations.

At the bottom of this page is a flow diagram of the relations between all the files and programs of FARS. It is also a clickable map for easy access of further descriptions and sample files. But first...

The FARS core program requires five input files:

object.nsl PIP representation of an object to be grasp
network.rsnet network wiring, weight matrices between layers
network.nsl default setting of the network and experiments, paramters of neural dynamics of different layers
protocol.nsl experiment protocol
track.trk names of units whose output would be monitored during the simulation

Two of the input files (object.nsl and network.rsnet) are created by some supplementary programs (mknet and genmcx). These supplementary programs are implemented in Perl. They themselves require input files that specify the type and high-level behaviour of the neurons in different layers and the set of grasps and objects to be encoded in the network.

How the input files are loaded into FARS is described in exp.script, which is parsed to the network. In return, the network generates four output files:

track.out.dump time profiles of the state of the units specified in track.trk
events.evn occurrence time of major events in the experiment
exp.log normal output trace of running FARS
synthetic.pet synthetic pet measurement during the simulation

The relations between all these files and programs are summarized in the following clickable map. To access more details or inspect sample files, simply click on the icon of interest.

FARS files