Dendronic Decisions Limited

Advanced Computing Research -- Expertise in Machine Learning

ALNfit Pro

Above is an actual screen shot after a run.  These are the main features of ALNfit Pro:

• select any column of the data file as the output variable, the others are input variables by default

• define input variables using any columns from the data file

• define lags on any input variables for multiple time series analysis

• force monotonicity of the output with respect to any input, or more generally put bounds on any weight

• don't waste time looking for good parameters -- training is fully automated, network grows as needed

• use importance values of variables to remove less important ones for the next run

• control generalization by setting a single number from 1 to 10 (number of ALNs averaged in bagging)

• organize setups and result files of many runs using the time as a prefix on file names (e.g. 1705 for 5:05 PM).

• use the "Options" button to choose alternate validation and test sets, or to optimize for real-time use

• ALNfit Pro is free!  Click to download  ALNfit Pro.exe or   ALNfit Pro.zip

• The Dendronic Learning Engine and ALNfit Pro are now OPEN SOURCE under the Lesser Gnu Public License (LGPL).
  The code incorporates parts using the Boost and QuantLib Licenses.  Programmers who know C++ are invited to work on applications or enhance the Dendronic Learning engine SDK. The present plan is to put this on SourceForge after some initial cleanup.

•   Click here to download libaln.zip (<30 MB). Click here to download some files for testing ALNfit Pro (<500 KB)

Send mail to with questions or comments about this web site.
Copyright © 2005 Dendronic Decisions Limited. All rights reserved.
Date Modified: February 18, 2005