manpages

ff_bignet(1)

ff_bignet(1) SNNS ff_bignet(1)

NAME

ff_bignet - complex network generator for the neural network simulation package SNNS

SYNOPSIS

ff_bignet plane-definition ... link-definition ... [ output-file ]

plane-definition
defines the various layers of the network. There must be at least one plane definition, but there might be any number given on the command line. Note, that you might have to define several planes for a single network layer, if you want to define a more complex network connection pattern than just "full-connection". One plane definition takes the form

-p x y [ act [ out [ type ] ] ]

x : number of units in x-direction

y : number of units in y-direction

act : optional activation function. Default: Act_Logistic

out : optional output function. Default: Out_Identity.

type : optional layer type. Valid types: input, hidden, or output

Default values : 1st plane: input; last plane: output; all others: hidden

link-definition

defines the connectivity of the network. There must be at least one link definition, but there might be any number given on the command line. One link definition takes the form

-l source-section target-section

Both source and target section definitions take the form

p cx cy cw ch ux uy mx my

p : number of the plane to connect

cx : x position of cluster in plane p

cy : y position of cluster in plane p

cw : width of cluster in plane p

ch : height of cluster in planep

ux : x position of a distinct unit in plane p

uy : y position of a distinct unit in plane p

mx : delta x for multiple fields in plane p

my : delta y for multiple fields in plane p

Unused parameters have to be specified as 0. A series of 0s at the end of each section may be abreviated by a '+' character.

OPTIONS

output-file

name of the generated network file. Default is "SNNNS_FF_NET.net"

DESCRIPTION

ff_bignet can be used to automatically construct complex neural networks. The synopsis is kind of lengthy, so when networks are to be constructed manually, the graphical version included in xgui(1) is preferrable. The parameters to this tools are derived from the input fields of the bignet panel in xgui(1) If, however, networks are to be constructed automatically, e.g. a whole series from within a shell script, this program is the method of choice. Note that even though the tool is called ff-bignet, it can not only construct feed-forward but also recurrent networks.

EXAMPLES

ff_bignet -p 1 10 -p 1 5 -p 1 6 -l 1 + 2 + -l 2 + 3 +

creates a simple 10-5-6, fully connected feedforward network.

ff_bignet -p 1 10 -p 1 5 -p 1 6 -l 1 + 2 + -l 2 + 3 + -l 1 + 3 +

creates the same network including shortcut connections.

ff_bignet -p 1 10 -p 1 5 -p 1 6 -l 1 + 2 + -l 2 + 3 + -l 2 1 1 1 1 0 0 0 1 2 1 1 1 1 0 0 0 1

creates the same network including additional self recurrent links in the hidden layer.

ff_bignet -p 6 20 -p 1 10 -p 1 1 -l 1 1 1 6 10 + 2 1 1 1 10 -l 2 + 3 1 1 1 1 +

defines a network with three layers. A 6x20 input layer, a 1x10 hidden layer, and a single output unit. The upper 6x10 input units are fully connected to the hidden layer, which in turn is fully connected to the output unit. The lower 6x10 input units do not have any connections.

ff_bignet -p 1 2 -p 1 3 -p 1 1 -l 1 + 2 + -l 2 + 3 + -l 2 1 1 1 1 0 0 0 1 2 1 1 1 1 0 0 0 1

defines a 2-3-1 network, fully feed-forward connected with additional self recurrent links in the hidden layer.

SEE ALSO

The SNNS user manual under the SNNS installation directory or the hypertext manual at http://www-ra.informatik.uni-tuebingen.de/SNNS/UserManual/UserManual.html

snns(1) description of the SNNS package

AUTHOR

Guenter W. Mamier