ffnet.ffnet

Package ffnet :: Module ffnet

[show private | hide private]

Module ffnet.ffnet

Module containing ffnet class and utility functions.

Classes
`ffnet`	Feed-forward neural network main class.

Function Summary
	`exportnet(net, filename, name, lang)` Exports network to a compiled language source code.
	`imlgraph(arch, biases)` Creates special layered network where outputs are independent from each other.
	`loadnet(filename)` Loads network pickled previously with 'savenet'.
	`mlgraph(arch, biases)` Creates standard multilayer network with full connectivity list.
	`readdata(filename, **kwargs)` Reads arrays of numbers from ASCII file.
	`savenet(net, filename)` Saves network to a file using cPickle module.
	`tmlgraph(arch, biases)` Creates multilayer network full connectivity list, but now layers have connections with all preceding layers (not only the first one)

Function Details

exportnet(net, filename, name='ffnet', lang='fortran')

Exports network to a compiled language source code.
Currently only fortran is supported.

There are two routines exported. First one, for recalling the network,
is named as indicated by keyword argument 'name'. The second one,
for calculating partial derivatives, have the same name with 'd'
prefix. 'ffnet' and 'dffnet' are exported at default.

NOTE: You need 'ffnet.f' file distributed with ffnet
      sources to get the exported routines to work.

imlgraph(arch, biases=True)

Creates special layered network where outputs are independent from each other. Exemplary architecture definition: arch = (3, [(4,), (), (6, 3)], 3). With such an arch, imlgraph builds three independent multilayer graphs: 3-4-1, 3-1, 3-6-3-1 and merges them into one graph with common input nodes.

Simplified version of the above architecture syntax is: arch = (3, 3, 3) #exactly as in the mlgraph Three nets: 3-3-1, 3-3-1, 3-3-1 are merged in this case.

loadnet(filename)

Loads network pickled previously with 'savenet'.

mlgraph(arch, biases=True)

Creates standard multilayer network with full connectivity list.

readdata(filename, **kwargs)

Reads arrays of numbers from ASCII file. Accepts all scipy.io.read_array keyword arguments. Read documentation of read_array function for details.

savenet(net, filename)

Saves network to a file using cPickle module.

tmlgraph(arch, biases=True)

Creates multilayer network full connectivity list, but now layers have connections with all preceding layers (not only the first one)

Generated by Epydoc 2.1 on Tue Oct 23 19:15:38 2007

http://epydoc.sf.net