NeuralWorks Professional II/PLUS v5.5

Category Intelligent Software>Neural Network Systems/Tools

Abstract NeuralWorks Professional II/PLUS is an advanced professional neural network (NN) development system. Product allows you to build, train, refine and deploy neural network solutions. Its advanced features include:

InstaNet facility - This menu lets you select from any of the major network types to create any of the 28 major NN paradigms and dozens of variations supported by NeuralWare. Each major class of networks has its own custom menu with options and features specific to that class. For example, the Back-propagation Builder supports standard, cumulative, normalized cumulative, Delta-Bar-Delta, Extended-Delta- Bar-Delta, Quickprop, and Maxprop learning rules, as well as Cascade architecture and the Logicon Projection Network. Other major classes of networks include Learning Vector Quantization, Self-Organizing Map, General Regression Network, Genetic Reinforcement Learning, Probabilistic Neural Network, Radial Basis Function Networks, Counter Propagation, Modular Neural Network, Recurrent Networks, Reinforcement Learning, Fuzzy ARTMAP, and dialogs for special purpose networks or networks of historical significance (Adaline, Bidirectional Associative Memory, Brain-State-in-a-Box, Boltzman, Directed Random Search, Hamming, Hopfield, Madaline, Perceptron, Recirculation, and Spatio-Temporal Pattern Recognition).

FlashCode - Instantly create ANSI standard C code to deploy your fully trained network. FlashCode now supports all of the major network types. This does Not require any additional hardware. You can use these routines to deploy your application, or combine networks together with other advanced technologies such as Expert Systems and Fuzzy Logic.

Note: Contact NeuralWare for licensing details.

SaveBest - Prevent over-training with SaveBest. At user-specified intervals, the network is tested and the best-performing network is saved. Performance can be based on R-squared, Root Mean Square (RMS) Error, Average Classification Rate or custom designed performance measure. This feature lets you train the network to the optimum degree based on the criteria, which is right for your application.

Optimize the Hidden Layer - An enhancement to SaveBest for non- linear feed-forward networks (Backpropagation) provides the capability of automatically optimizing the size of the hidden layer. Each hidden unit is ranked based on the degree to which it improves the objective function. Those units with a negative or minimal contribution are eliminated. This capability can be used in conjunction with the facility to periodically prune a network of small weights.

ExplainNet - This is a unique facility that tells you why a neural network made its decisions. It shows you which inputs to the model have the most impact on the network output. This is vital for applications like credit scoring. Together with utilities available from NeuralWare technical support, it can be used to do variable selection for building more robust models.

Diagnostics - A wide variety of diagnostic tools to monitor network performance and diagnose problems are standard. Each of the major network types have specialized instruments, which can be instantly created and used to observe specific activity. More general instruments can be created with NeuralProbe. Custom instruments can be created using User Input/Output (IO) extensions.

Graphical User Interface - No matter what platform you choose, you will find the same icon-based tool palettes and “mousable menus”. Neuralware's interface is the same in every version of the product, making it easy to migrate to more powerful computing platforms.

Batch Interface - User Control provides the capability for automating the neural network development process through a custom control interface with batch files. These control programs written in C can load, train, and test networks. You can customize the operation of Professional II/PLUS to the rigorous demands of your production neural network development.

Inter-platform compatibility - Not only does NeuralWare support a wide range of development environments, data files and network files are fully interchangeable between them. This lets work groups in non- homogeneous computing environments transparently share data and network files.

Flexible Architecture - You can customize the network architectures, develop training procedures, and substantially customize the operation of this product through the InstaNet Interface Description Language. (IDL) language. This is a built-in text-based scripting capability, which can be used to architect a variety of neural network architectures and features.

System Requirements

Intel® x86 Architecture

Windows 95/98/ME/NT/2000/XP: Pentium/Celeron class processor 64MB memory, 10MB disk, CD Reader

Linux

Pentium/Celeron class processor, 64MB memory, 10MB disk, CD Reader

Sun Microsystems - Sparc/UltraSparc Architecture

Sun Solaris 8.0 or greater, 128MB Memory, 10 MB disk, CD Reader

SiliconGraphics - 128 MB Memory, 10 MB Disk, CD Reader

Additional platforms are available. Contact NeuralWare for details.

Manufacturer

Manufacturer Web Site NeuralWorks Professional II/PLUS

Price Contact manufacturer.

G6G Abstract Number 30108

G6G Manufacturer Number 101900