havBpNet++ Demo - C++ classes for implementing Feedforward, Simple Recurrent and Random-Order Recurrent Neural Nets trained by Backpropagation.
NN Demos havBpETT - Screen Shots

havBpETT is a little Visual Basic application which demonstrates use of a DLL version of the havBpNet++ c++ Neural Network class library. The demo allows you to describe (or restore), train, save and consult a simple feed forward, recurrent or sequential neural network (trained by backpropagation).

We have prepared screen-shots to illustrate the 4 main user interface screens of the havBpETT Demo. below are shots of the...

  • Main Screen - A toolbar type screen from which main actions are selected.
  • Network Control - The interface for describing the network layers, connections, etc.
  • Layer Description - The interface for defining certain layer parameters like number of nodes, learning-rate, activation-function, error function etc.
  • Training Control - The screen used to start and stop training, change certain network parameters and watch training progress.

The Main screen presents both a toolbar and an information display. The toolbar may be used to select overall actions (such as data or network control). Both buttons and menu items are provided for all main actions. The information display presents a summary of certain layer and network parameters. Also provided is an information display line in which messages will appear as the cursor is placed on various buttons and fields.

As delivered, the 1st UserButton is assigned to the Notepad utility. You may reassign this UserButton and make assignments to the remaining UserButtons to allow easier transition between havBpETT and other applications, such as spreadsheets and plotting packages.

Main Screen

The Network Control screen is used to describe the overall network's configuration. Connections between layers are enabled/disabled by clicking on the connectors located under a layer. Inactive connections are dashed light-gray lines. Active connections are color coded according to their type: Forward, Weighted-Copy Recurrent and Random-Order Recurrent.

Layer parameters are entered for each layer individually. A layer is selected by clicking on the layer.

The Network Control screen also contains the controls for Restoring a previously saved network into memory and for Saving the current network to a save-file.

Connecting the layers

The Layer Description screen allows you to set various layer parameter values.

When a layer button in the Network Control screen is clicked, a sub-screen is opened that allows the user to set/select the values of various parameters associated with the layer. Examples of these parameters are layer-size, learning-rate (Beta), momentum (Mu), etc.

By selecting the Accept button, the current parameter values are communicated to the system. By selecting the Cancel button, any changes in the layer's sub-window are discarded and the original values are used.

Set Layer Parameters

The Training Control screen is opened when you click the Train button on the Main screen (Wand-in-Hand).

In this window, you can start and stop training of the current network. You can control the error-mode (Analog or Sign-Only) and you can specify that the network should be saved to disk each time the network's performance is better than it has yet been.

On the Training window, you can change certain network and layer parameters by selecting the appropriate item, entering the new value and hitting the Submit button. If training is currently in-progress, then the new values will take effect on the next training pass through the data.

Both a digital and graphical display of network performance is presented toward the bottom of the Training window. The Graph shows the %-error and may be turned on/off as desired. Turning the graph OFF will noticeably increase training speed for smaller nets with relatively small training-data sets. When the graph is off, it is still updated so, if you turn it back on, you will see the results of passes when it was off. You are also allowed to specify whether the error-axis is displayed with linear or logarithmic scaling. Finally, you can print or save a copy of the graph at any time.

Training Control and Progress

[ HOME ]
Copyright © 1994-2017 by hav.Software and Horace "Kicker" Vallas. All Rights Reserved.

hav.Software havBpNet:J, havFmNet:J, havBpNet++, havFmNet++, havBpETT, havCNet, WebSnarfer, havIndex and havChat are all trademarks of hav.Software.

Java and all Java-based marks are trademarks or registered trademarks of Sun Microsystems, Inc. in the U.S. and other countries.

There may be other trademarks or tradenames listed in this document to refer to the entities claiming the marks and names or products. hav.Software disclaims any proprietary interest in any trademark, tradename or products other than its own.

6,678,884 / 31,565,783
Page Modified Mon May 23 18:23:24 CDT 2011
TennisPensacola.com - the official online hub for tennis in Pensacola, Fl. Kick-kick the  Australian Shepherd - Welcome the newest Vallas pack member In memory of Bear Dudley Vallas, a Long Time and Faithful Companion ndex of Kicker's Classical Guitar Videos Bonsai Information, sites and icons etc. Horace's Horror - just why is it so hard to test software?