Nenet

survey performed by Juha Ikonen, May 21st 1999

Disclaimer: If any information on this page simply is not true, please tell us about it and we'll correct it ASAP.

Disclaimer: The opinions and observation herein should be considered personal of the person having performed by the survey, at the time of the survey. They do not reflect any official standing of his employer, of the Laboratory of Computer and Information Science or the Neural Networks Research Center.

General

Program name	Nenet 1.0b
Availability	Available freely from the website http://koti.mbnet.fi/~phodju/nenet/Nenet/General.html
Purpose	Educational
Operating system	Windows 95, Windows NT 3.5x and Windows NT 4.0
User interface	Windows 95 - style graphical user interface. [speed of usage,user friendliness]
Documentation	Good online help which also reveals some theory of the SOM algorithm

[General comments]

SOM features

map parameters
Teaching algorithm	Standard, map is updated after presenting each vector of the training data set. Implementation appears to be correct.
Map size	Minimum 2 x 2 cells, maximum 100 x 100 cells
Map lattice and shape	Map lattice is rectangular or hexagonal, map shape is always rectangular.
Neighborhood function	Function type: Bubble or gaussian
	Neighborhood size (h): Both initial and final values can be set by user. [type (linear, 1/t, other), parameters]
	Learning rate (alpha): Linear or inverse time (1/t). [type (linear, 1/t, other), parameters]
Initialization	Linear or random.
Distance function	Euclidian
Unknown components	Not allowed
Teaching length	Explicit, measured in steps. Ending conditions can not be used.
efficiency
Speed [Windows NT 4.0, 200 MHz Pentium MMX, 128 MB RAM]	16 seconds for standard run.
Results	Normal [quantization error, topographic error]

[Comments on SOM implementation]

Usability

preprocessing
Input formats	ASCII for both of the data and map files
Data handling and selection	Scaling by Range: the values are scaled between [0,1] for each component separately. Scaling by Variance: the variance of the components is scaled to 1 and the mean to 0 for each parameter level. No data selecting functions, the program processes the entire data set. [flexibility,usability]
postprocessing
Output formats	ASCII
Map measures	Not implemented.
Labeling	Simple, neurons and vector components can be labelled.
Clustering	By visualization.
visualization
Inspection of neurons	Simple, vector values can be viewed by selecting desired neurons.
Clusters/map shape	Standard and interpolated Umatrix, 2D and 3D histograms.
Correlations	Not implemented.
Data projections	A group of data vectors can be projected on a map (histogram presentation).
Markers	Labels only.

http://www.cis.hut.fi/projects/somtoolbox/links/nenet.shtml
somtlbx@mail.cis.hut.fi
Monday, 09-Oct-2000 12:53:09 EEST