NeuNet Pro Program Screens
Table Of Contents

Screens

Configure Project

Advanced Config

Data Split

SFAM Train

SFAM Browse

Confusion Matirx

BackProp Train

BackProp Browse

Scatter Graph

Time Series Graph

Advanced Project Configuration

Number of Hidden Nodes User Min/Max Field Name Field Type Field Statistics OK
This screen allows advanced users to override the default settings for User Min/Max and the number of nodes in the BackProp hidden layer .
For most projects, the default values work very well, so use of this screen is optional.
Warning: Any changes on this screen will cause a loss of all previous training for this project.

Number of BackProp Hidden Nodes:
  • The optimal setting for this value depends upon how many input fields you are using and how well the prediction value is related to the input values.
  • The default suggested value is 2 x SquareRoot(Number_Of_Input_Fields + 1).
  • If too many nodes are used, the neural net will be slower to train and it will tend to "memorize" instead of "learn". The accuracy will improve on the training set, but might be worse on the testing set. The chance of discovering spurious correlations will be increased.
  • If too few nodes are used, the neural net will train more quickly, but the accuracy on the training set may decrease. The accuracy on the testing set may improve. The chance of discovering spurious correlation will be reduced.

User Min/Max:

  • User Min/Max for each field defaults to +- two standard devisations around the average value for this field in entire data table (version 2.3).
  • User Min/Max is used to control the internal "normalization" of the data:
    A data value greater than or equal to User Max is normalized to 1.0
    A data value less than or equal to User Min is normalized to 0.0
    All other data is proportioned between 0 and 1.
    For example a value half way between User Min and User Max is normalized to 0.5
  • In some cases, you may wish to adjust User Min/Max:
  • If you expect future data to occur outside the range of the Min/Max for your current data range, it is helpful to widen User Min/Max to reflect the range of future data values.
    For example, if all your fields are data readings from the same instrument, you may wish to set User Min/Max to the range of that instrument.
  • If your data contains a few outliers, these outliers will distort the normalization for the rest of your data. You should narrow the User Min/Max so the normalization is scaled to the majority of your data.
    For example, suppose you are working on 1000 rows of data and 997 rows have a certain field valued between 100 to 200. Three rows have this field valued at 1000. You should decrease User Max from 1000 to around 200, so the normalization of the data reflects the majority range.
  • If you have a high portion of your data occurring at User Min and/or User Max, it is helpful to widen the range of User Min/Max so there is some "elbow room" for your data. This technique will tend to move your data out of the unresponsive tails of the sigmoid function, leading to better predictions.
    For example, suppose you are trying to predict a field that has values of 0 or 1. Try setting User Min to -0.25 and User Max to 1.25, and your predictions should improve.

Field Names:

  • The field names shown are extracted from the Ms-Access format of your data table.

Field Types:

  • The field types shown are extracted from the Ms-Access format of your data table.
  • NeuNet Pro internally treats all numeric and date fields as single precision real numbers (7 significant digits).
  • Text fields remain as text, which can be used only for SFAM predictions.

Field Statistics:

  • This report show the actual Minimum, Maximum, Average, and Standard Deviation for every numeric field across your entire data table.
  • Use Print to grab a hard copy of this interesting report.

OK:

  • The OK button returns you to the Configure Project screen.
  • If you make any changes during a visit to the Advanced screen, you will not be able to Cancel out of the Configure Project screen because you have already changed the project configuration.



A Complete Neural Network Development System

CorMac Technologies Inc.
34 North Cumberland Street ~ Thunder Bay ON P7A 4L3 ~ Canada
E m a i l