MemBrain

Category Intelligent Software>Neural Network Systems/Tools

Abstract MemBrain is an advanced graphical neural network editor and simulator for Microsoft Windows, supporting neural networks of arbitrary size and architecture. Products features/capabilities include:

Mathematical Model used by MemBrain -- The neuron and link model of MemBrain is very flexible: Everything from simple time invariant Feed- Forward Networks to Nets with arbitrary loopback connections and signal run-time delays on the links can be simulated.

It is even possible to connect the output of a neuron to its own input. In principle every net in MemBrain is a valid net. Through the consequent object oriented approach every neuron and also every link in MemBrain can have different properties. Links can have user-defined logical lengths, so that real run-time behavior can be simulated and visualized.

Learning algorithms that are currently implemented in MemBrain --

Supervised learning:

1) Standard Backpropagation (only forward links are trained);

2) Standard Backpropagation with momentum (only forward links are trained);

3) Backpropagation with support for loopback links;

4) Backpropagation with support for loopback links and with momentum;

5) RPROP (Resilient Backpropagation) with support for loop-back links;

6) Cascade Correlation with support for loopback links (using Backpropagation with momentum);

7) Cascade Correlation with support for loopback links (using RPROP).

Unsupervised learning:

1) 'Winner Takes it All' for SOMs ('Self Organizing Maps').

Automatic Network Analysis -- You don't have to tell MemBrain what Net architecture you're currently building. Just place as many neurons as you like and where you like and connect them in the way you like. Besides the possibility of interconnecting neurons by drawing single links there are also some very good editing features available in MemBrain that allow insertion of thousands of links with just a few mouse clicks.

The integrated 'neural network analysis' feature detects the logical structure of the net, such as layers and loopback links, whenever needed automatically in the background. This information can also be visualized.

Open TCP/IP interface -- MemBrain incorporates the possibility to connect to other MemBrain instances or MemBrain compatible device drivers over a Transmission Control Protocol/Internet Protocol (TCP/IP) link. It doesn't make any difference if the other applications are running on different computers or on the same machine.

MemBrain's TCP/IP Web link protocol is provided as part of the context sensitive HyperText Markup Language (HTML) help that comes with the application. Anybody who wants to connect certain components to MemBrain can implement a TCP/IP based service that can communicate with the target on one end and with MemBrain on the other end and thus serve as Input/Output (I/O) "driver" for MemBrain.

Simple Scripting Language -- MemBrain features a simple scripting language which allows automation of often needed procedures like loading, training, validating and saving a net. Furthermore many other operations and settings in MemBrain are accessible through script commands. Thus, MemBrain scripts also provide a means for remotely controlling MemBrain from other software systems or for preparing small presentations that show the live operation of a neural net.

The MemBrain DLL -- Many functionalities of MemBrain are also available as a Dynamic Link Library (DLL) version.

With the MemBrain DLL you can incorporate neural nets that have been designed with MemBrain into your own applications, No matter what programming language you use.

The DLL features include:

1) Loading an arbitrary number of neural nets from MemBrain files;

2) Assigning values to the input neurons of the neural nets;

3) Performing simulation steps of the neural nets;

4) Reading back the results from the output neurons of the neural nets;

5) Saving the neural nets back to a file;

6) Additionally, all functionalities, to train the nets, to create, load, import and export data sets and much more are available through the DLL. Thus, neural nets can now be re-trained within your own applications.

By this means, it is possible to incorporate neural nets from MemBrain into a production system without the need of having MemBrain running in parallel and without having to deal with the TCP/IP interface or the scripting language of MemBrain.

Additional features/capabilities of MemBrain are:

1) Comprehensive, context sensitive HTML-Help including a 'Getting Started' tutorial.

2) Import and export of raw data is possible via plain Comma Separated Values (CSV) files.

3) New: Automatic Split Off of validation data sets.

4) New: 'On-The-Fly'-Validation - already possible during training using separate validation data.

5) New: Graphical representation of target/actual data in real-time during training.

6) New: Supervised learning, now also possible with incomplete data sets.

7) Opening files when neurons fire (e.g. executable or batch files).

8) Single step mode or continuous simulation with adjustable timing.

9) Configurable export functions for neural nets.

System Requirements

Manufacturer

Manufacturer Web Site MemBrain

Price MemBrain and the MemBrain DLL are free for private and for non-commercial use (e.g. in universities for educational purpose). There are no restrictions with respect to functionality. Licensing.

G6G Abstract Number 20138

G6G Manufacturer Number 102711