Key Features

 

Data Processing
  • Import Excel files

  • Import CSV and TXT files

  • Import bitmap image files

  • Import binary files

  • Extensive pre-processing facilities

  • Date and Time encoding

  • Integer, real, boolean, text or image modes

  • Many methods of handling missing values

  • Min/Max column values for scaling

  • Outlier handling

  • Random and manual data partitioning

  • Data subsets

  • Check for duplicate rows

  • Column value classification

  • Range extender and filler

  • Shuffle rows

Building the Neural Network
  • Input and output selection

  • Multiple inputs and outputs

  • Check rows and columns are suitable to build a network

  • Automatic or manual production of hidden layers

Control Training and Validating
  • Automatic or manual learning rate and momentum

  • Automatic decay of learning rate and momentum

  • Global or independent input and output validating rules

  • Scoring rules

  • Automatic or manual network reconfiguration while learning

  • Stop after fixed number of cycles

  • Node and Weight freezing

  • Variable validating periods

  • Fixed time stop

  • Variable speed learning for visual demonstrations

  • Many methods of early stopping

  • Validating correct or within range

  • Jitter and Noise

  • Random and Balanced presentation

Special Files
  • Save any part of the network to TXT or CSV files

  • Save the data grid to TXT or CSV files

  • Save learning progress to TXT or CSV files

  • Save backup while learning

Auto Save while learning
  • Variable save period

  • Save when error reduces

  • Save when validating results improve

 

Macros and Scripts
  • Record and Play macros

  • Extensive script language

  • Add commands and scripts to macros

  • Run scripts from the command line or other applications

  • Single step macros

  • Run background scripts while hidden

Querying
  • Query trained networks manually

  • Query networks from external files

  • See output values change when changing input values

  • Seek high or low outputs

  • Cycle seek though all inputs

  • Save results to TXT or CSV files

  • Query inputs can be extended beyond the training range

  • Extrapolated results can be produced

Forecasting
  • Forecast future values with multiple networks

  • Allow forecasts to extend beyond training limits

  • Assess risk of forecasts

  • Restrict forecasts to upper or lower training limits

Associations and Clusters
  • Automatically find associated inputs and outputs

  • Find inputs and outputs that form clusters

  • Save associations and clusters

  • Build networks from clusters

Leave Some Out Validating
  • Sequential leave out subsets selection

  • Random row selection validating

  • Shuffle before validating

  • Random leave out subsets selection

  • Multi-fold cross validating

  • Comprehensive report

Node Reduction and Weight Pruning  
  • Prune insignificant weights while learning

  • Reduce network to minimum size

Views  
  • Data grid

  • Network nodes and connections

  • Learning progress graph

  • Column values graph and trends

  • Actual and predicted outputs for training and validating examples

 

  • Training and validating examples errors

  • Input importance and relative importance

  • Input sensitivity and relative sensitivity

  • Input and output associations

  • Diagnostic array

  • General information

Help system, online FAQ and user manual
 

Free support for all users

 

Many ready trained samples

 
Copyright 2002 - 2007 Neural Planner Software
Last Updated:  4th June 2008