Neural Lab

Neural Lab is a free, neural network simulator that designs and trains artificial neural networks for use in engineering, business, computer science and technology. It integrates with Microsoft Visual Studio using C (Win32 - Wintempla) to incorporate artificial neural networks into custom applications, research simulations or end user interfaces.

It provides a visual environment to design and test artificial neural networks.

The latest Neural Lab version is 4.1. The two major versions are version 3.1 and 4.0.

Version 3.x

Version 3.1 is navigated using a standard computer mouse. Version 3.1 is considered easier to use, however, it is difficult to perform complex tasks programmatically. Version 3.1 is therefore primarily useful for people without a programming background.

The version 3.1 tutorial provided very little theoretical background on artificial neural networks. Despite the number of examples, most of the examples focus only on multi-layer networks with supervised training.

Version 4.x

In version 4.0, it is possible to perform neural network operations by writing code. The code is very similar to C/C++, Java or C#.

In version 4.0, the authors incorporate background information on artificial neural networks.

Version 4.0 incorporates Kohonen networks that can be trained without supervision and probabilistic neural networks.

Features

The tools allow reviewing and analyzing the structure of the training set.
The activation of the neurons for each case in the data set are visible. The tutorial provides examples in prediction, data mapping, data classification and autoassociative memory problems.
Once a network has been trained, it is possible to save it to a file. The file can be opened using Microsoft Visual Studio to create a standalone application that can employ the network.

Applications

Specific examples of neural networks include:

Prediction
Mapping
Auto Association
Classification
Network Simulation

Wintempla

Neural Lab is developed using Wintempla (a plug in that works with Microsoft Visual Studio). Wintempla encapsulates Win32 and simplifies the development of Microsoft Windows applications using C++ and native Win32 APIs.

Wintempla is a tool that integrates with Microsoft Visual Studio. Wintempla encapsulates Win32 to simplify the creation of Web and Desktop applications using C++ and object-oriented programming. The programmer has the option to use the native Win32 APIs or the Wintempla classes.

Wintempla includes:

A tutorial with key concepts in programming
Videos to illustrates how common control instructions (such as: if, else, for, while, etc.) work
Many examples and problems that can be used in: programming classes, SQL, PLSQL, Graphics
Support to create SQL database applications
SQL Import to create (in seconds) desktop or web applications from a SQL file
Simulated annealing optimization
Genetic algorithm optimization
Asynchronous module for Digital to Analog converters (DAC)
Asynchronous module for Analog to Digital converters (ADC)
Asynchronous module for serial ports
Multithread applications
Document printing
Microsoft Windows services
GUI deployment
Digital Signal Processing (remez, FFT and Filtering)
Common Object Model (COM)
A Lexical Analyzer, a compiler and virtual Machine
Artificial Neural Networks
Matrix operations
Data Visualization: Pie Chart, XY Chart, Polar Chart, Histogram, 3D Visualization, Simulation View
Native support for string manipulation using the STL
Native support for Math operations using the STL
Native support for data file storage
GDI Game application
Support for DirectX applications
Support for Open GL applications
Support to create PDF files programmatically

File extensions

.lab Neural Lab code (a UNICODE text file)
.lay A multi-layer neural network file
.lax A complex-domain multi-layer neural network file
.koh A Kohonen neural network file
.prb A probabilistic neural network file
.csv A comma separated data file

References

Masters, Timothy (25 July 1994). Signal and Image Processing with Neural Networks: A C++ Sourcebook. John Wiley & Sons. ISBN 978-0-471-04963-0.
Masters, Timothy (17 April 1995). Advanced algorithms for neural networks: a C++ sourcebook. Wiley. ISBN 978-0-471-10588-6.

External links

This article is issued from Wikipedia - version of the Thursday, February 04, 2016. The text is available under the Creative Commons Attribution/Share Alike but additional terms may apply for the media files.