Changes Between Release 1.1 and 1.5

Top  Previous  Next

 

This section will detail changes between NeuroShell 2 Release 1.5 and Release 1.1.

 

Speed

NeuroShell 2 trains twice as fast as did previous releases of NeuroShell 2, assuming graphs and statistics are off.

 

File Changes

The file format for the .FIG, .MMX, and .DEF files has been changed in NeuroShell 2 Release 1.5.  If you created a problem in a previous release of NeuroShell 2, you should use the Define Inputs/Outputs module to create a new .MMX file and use the Design module to create a new .FIG file.

 

Release 1.5 .DEF files are not compatible with release 1.1 .DEF files.  If a user has old .DEF files he/she does not want to retrain under Rel 1.5, then these will have to be called through the old DLL (which was renamed NSHELL2.11 by the Rel 1.5 SETUP program; the extension refers to release 1.1).  To do this, the OpenNet, FireNet, etc. prototype statements will have to be changed to reference "NSHELL2.11" instead of "NSHELL2.DLL", but only in programs referencing these old .DEF files.  If you're using EXCEL, the CALL statement needs to refer to "NSHELL2.11" also.

 

If your data file includes an * in a cell beneath a column labeled A (Actual output), the * will be replaced with a 0 and prediction will be made in that row when you apply a network.  A prediction will not be made in a row if your data file includes an * in a cell beneath a column labeled I (Input).  (The column labels were specified in the Define Inputs/Outputs module.)  Previous releases of NeuroShell 2 would not apply a trained network to a data row if it contained an * in either an A or I column.

 

The Papyrus Clarity file import module was removed.

 

 

Initial Weights

The initial weights for networks created in NeuroShell 2 Release 1.5 are different than those in previous releases. If you are trying to reproduce a problem in Release 1.5, you may get slightly different results than you did in a previous release.

 

Architecture and Parameters

The default number of hidden neurons for a 3 layer network is computed with the following formula:

# of hidden neurons = 1/2 (Inputs + Outputs) + Sqrt(# of Patterns)

 

For more layers, divide the number above by the number of hidden layers or slabs.

 

We have found this formula to be better than the default formula in NeuroShell 1, which was:

# of hidden neurons = 2 * square root (number of inputs or defining characteristics + the number of outputs or classifying characteristics) rounded down to the nearest integer

 

Market Indicator Package

The Edit Menu now has the Display Variable in Help Bar Option which displays the entire text for a variable in the Help Bar at the bottom of the screen.

 

The Optimal Menu has added two new selections:

 

Multiple Variable Mode: Selecting this mode allows you to add Optimal Indicators to the single variable that is highlighted in the Variable Selection box and to all of the subsequent variables in the list (except for lead variables).

 

Multiple Lead Mode: Selecting this mode allows you to add Optimal Indicators to the first variable that contains the word "Lead" and to all subsequent variables that contain the word Lead.

 

The Multiple Variable Mode and Multiple Lead Mode may be used independently of one another or in combination.

 

Tutorial Example

A simpler tutorial example has been added to the Help file.  Refer to Tutorial 1 for more information.