attrasoft

Attrasoft

Attrasoft PolyNet 
for Windows 95/98
Version 4.0 (7/1999)
About

New

Image Search Engine

Products

Services

Order

Shareware/Demos

Future Products

Links

Search This Site


All Attrasoft products are Y2K Compliant!

1. Introduction
1.1 Neural Networks
1.2 Hopfield Model and Boltzmann Machine
1.3 History of PolyNet
1.3.1 ABM
1.3.2 Attrasoft Neural Applications
1.3.3 PolyNet
1.4 Training and Recognition
1.5 Neural Network System
1.6 PolyNet's Version
 

1.   Introduction

Biological neurons are believed to be the structural constituents of the brain. A neural network can:

  • learn by adapting its synaptic weights to changes in the surrounding environments; 
  • handle imprecise, fuzzy, noisy, and probabilistic information; and
  • generalize from known tasks or examples to unknown ones.
Artificial neural networks (ANN) are an attempt to mimic some, or all, of these characteristics. 

A neural network can learn. One of the neural net paradigms is supervised learning. In supervised learning, a neural net directly compares the network output with a known correct or desired answer in the training process. 

1.1   Neural Networks

Neurons are living nerve cells that come together to form complicated biological neural networks. A typical human brain has 100 billion neurons. Each neuron is connected to 1,000 to 10,000 neurons. The human brain contains approximately 100 to 1,000 trillion interconnections. 

An artificial neural network (ANN) is either software or hardware that can simulate biological neural networks. Artificial neural networks can perform computations and have learning abilities. These features distinguish neural network software from other software.

When you bought PolyNet, you bought software that simulates a neural network. This is a blank neural net and there is nothing stored in the network. Your job is to train the network. After that, the PolyNet is ready to serve you.

A neural network is characterized by:

  • network topology, 
  • connection strength between neurons,
  • node properties,
  • internal controls, and 
  • the updating rules. 
In PolyNet, the neuron states are 0, 1, 2, .... For simplicity, the PolyNet will use decimal neurons, i.e., the neuron states are 0, 1, 2, ... 9. If you want a different set of neuron states, a customized PolyNet has to be ordered from Attrasoft

To train a network, training data is provided to the network. Training data consists of many patterns like this:

00100
01100
00100
00100
00100
00100
01110.
This is an image of a 5x7 character, '1'. These training patterns will influence the network. Every time the network looks at a training pattern, the network stores the information by modifying the neuron synaptic connections. Modifying the values of the connections represents a learning process: the neural networks learn their environment by changing their internal connections. After a while, these synaptic connections hold certain values. These values represent the neural network's memory and it can be used to perform certain tasks.

1.2   Hopfield Model and Boltzmann Machine

One of the most popular networks is the Hopfield network. This recurrent net is completely connected. The Boltzmann Machine is closely related to the Hopfield model. The Boltzmann Machine is a special type of neural network, in which each neuron configuration has a certain probability to appear. (The name comes from the following fact: the Boltzmann Machine is a probabilistic neural network which forms a Markov chain; the invariant distribution of the Markov chain is similar to the "Boltzmann Distribution" in statistical physics).

Most software is "programmed" to perform certain tasks. These tasks are fixed. For example, chess game software will not play solitary. Neural network software is not programmed; it is trained. Neural networks do what they are trained to do. What you train is what you get.

1.3   History of PolyNet
 

Modern digital computers outperform humans in the domain of numerical computation and related symbol manipulation. However, humans can effortlessly solve complex perceptual problems, like recognizing a person in a crowd from a mere glimpse of his face, so quickly that it would dwarf the world's fastest computer.
 

1.3.1   ABM
To give the machine the ability to solve complex perceptual problems, Attrasoft developed its first generation neural software: ABM (Attrasoft Boltzmann Machine) for Windows 95/98. The first software, ABM 1.0, was developed in 1995.

ABM is a software simulation of the Boltzmann Machine. ABM is not "programmed" for a particular task. ABM has to be "trained" to perform certain tasks, therefore, before you can use the software, you have to train it. It is a medium size neural network, designed to operate between 1,000 to 1,000,000 external neurons (external neurons mean input and output neurons; a hidden neuron is not an external neuron).

ABM:

  • Nominated by PC Magazine for the 1996 Ziff-Davis Shareware Award;
  • Simulates two powerful types of neural networks, the Boltzmann Machine and the Hopfield Model; 
  • If it fails to recognize a pattern, one retraining will fix it;
  • If you train with the wrong information by accident or by design, retraining will correct it;
  • Can handle conflicting training data;
  • Supports 1D and 2D translation, scaling, and rotation- symmetries;
  • Works for any pattern recognition problem; 
  • Supports up to 10,000 external neurons;
  • Includes more than ten examples to walk you through, including one with more than 4,000 classes;
  • Learns more than 4,000 characters in 20 seconds and recognizes one of the 4,000 characters in 0.5 seconds, a speed unachievable by humans;
  • ABM does not lose focus or clarity as the net gets larger.
1.3.2   Attrasoft Neural Applications

ABM is the brain for all the first-generation Attrasoft software. The applications are:

Content-based Image Retrieval for local drive

Attrasoft ImageFinder for Window 95/98, v3.4, will do this. Application examples are:
  • Finger print recognition;
  • Signature recognition;
  • Handwriting zip code recognition;
  • ...
Content-based Image Retrieval for the Internet Attrasoft Internet ImageFinder, v3.3, will do this. Application examples are:
  • Internet Image Search. 
Content-based Image Search Engine for the Internet
Attrasoft ImageHunt, v3.5, will do this. 


Image Recognition

Attrasoft ImageClassifier 3.1 will do this. Application examples are:
  • Finger print recognition;
  • Signature recognition;
  • Handwriting zip code recognition. 
Time Sequence Prediction:  Attrasoft Predictor 2.6 will do this. Application examples are:
  • Predict Stock Market
  • Predict earnings and revenue of a company
  • Predict short term and long term interest rates
  • Predict commodity (gold, oil, corn,...) prices
  • Predict regional/worldwide price fluctuations for a particular merchandise item 
  • Predict various indicators/indexes of the economy for strategic thinking and policy issues
  • Predict Lottery Numbers 
  • Predict Dynamic Systems
  • Predict Markov Chains
  • Predict other complex systems
Function Inference Attrasoft PolyNet 2.5 will do this. Application examples are:
  • Decide whether to grant a loan to a customer, student *, ...
  • Decide whether to hire a person or not
  • Decide whether a patient has cancer *, heart disease *, thyroid disease *, ...., or not
  • Decide how likely a patient is going to get cancer, heart disease *...
  • Decide the classification of a virus/bacteria within a group of virus/bacteria *, ...
  • Decide how much a house *, a car *, a boat, ... , is worth
  • Decide whether your supervisors evaluate your employees fairly
  • Decide whether your system (electronics systems, computer network, retail store, sexual harassment situation) is healthy
  • Decide regional supermarket price predictability based on accumulated database, ...
  • Decide which career a child should choose 
  • Decide the outcome of a random event, like a horse race, ...
  • Identify an item in Forensic Science *, ...
  • ETC....
*Examples based on real data are in PolyNet 2.5. Sound Recognition
Will be available.


1.3.3   PolyNet

ABM simulates the binary Hopfield model and the binary Boltzmann machine. PolyNet simulates the polytomous Hopfield model and the polytomous Boltzmann machine. PolyNet is the first product in the second-generation. The first interface to be developed will be jpg/gif interface.

Figure 1. Attrasoft PolyNet.
 
 

PolyNet is derived from ABM. PolyNet uses decimal neurons instead of binary neurons. PolyNet will allow a maximum number of 10,000 decimal neurons to be used. There are two basic types of tasks we expect the software to perform:

  1. Classification: Given a pattern, find its class;
  2. Determine a Pattern: Given a classification and part of a pattern, complete the pattern.
    1.4   Training and Recognition
The ability to learn is a fundamental trait of intelligence. Neural nets can learn from examples. As a neural net simulator, PolyNet can also learn from examples. PolyNet is a blank artificial neural network. PolyNet has to be trained before it can perform even one task for you. There are two basic phases in using PolyNet:
  • Training; and 
  • Recognition.
In the training phase, data is imposed upon a neural network to force the network to remember the pattern of training data. A neural network can remember the training data pattern by adjusting its internal synaptic connections. 

In the recognition phase, a part of the input data is not known. The neural network, based on its internal synaptic connections, will determine the unknown part. 

A typical problem for a neural network to solve is the classification problem. The data is a set of doublets: (pattern, class). In the training phase, the network is taught which pattern belongs to which class. In the recognition phase, only the patterns are given to the network, and the network decides the classification of the patterns. Alternatively, when a part of a pattern and a classification are given, or just a part of a pattern is given, the network is asked to complete the pattern.
 
 

    1.5   Neural Network System
The neural network attempts to simulate the human brain. The human brain has front-end subsystems and can only handle preprocessed information. The front-end subsystems for humans are eyes, ears, ... The same is also true for neural networks. 

There are two types of data used by neural network systems: user data (or application data), and neural data. Neural networks use neural data. User data depends on the applications. To put it in another way: neural nets speak neural language, users speak user language. 

The information processed by a neural network has to be prepared by a front-end subsystem. This is called data encoding. A neural network can not usually handle the user-application data directly. Similarly, after neural computation, the result usually does not make sense to humans directly; the front-end system is responsible for converting the neural output data back into user-application data. This is called data decoding.

Front-end systems are basically a language translator. The Attrasoft has several interfaced packages:

  • Predictor 2.6
  • DecisionMaker 2.5
  • Internet ImageFinder 3.3
  • ImageFinder 3.4
  • ImageClassifier 3.1
  • ImageHunt 3.5
A neural network system consists of the front-end subsystem and a neural network subsystem:

Front-end subsystem

  • Data encoding: to convert application data to neural input data;
  • Data decoding: to convert neural output data back to application data.
Neural network subsystem
  • Classification: patterns are given to the network, and the network decides the classification of the patterns;
  • Pattern Fix: parts of a pattern and a classification are given, or just a part of a pattern is given, the network is asked to complete the pattern.
Examples of front-end subsystems:
  • JPEG Images to neural data;
  • GIF images to neural data;
  • Fax data;
  • Various databases;
  • Handwritten words;
  • Scanned documents;
  • Sound data;
  • Stock Market.
PolyNet is a neural network subsystem. It takes the text files as inputs and produces the text files as neural output. Customized software, which has the desired user interface, can be ordered from Attrasoft for any application.

To summarize, the neural computation process has three stages: Data encoding, Neural Computation, and Data Decoding. Data encoding and decoding are application-dependent, while the neural network is not application-dependent. PolyNet is a neural network simulator that is not application-dependent. It can be interfaced with any front-end systems to solve any problem. Attrasoft PolyNet is a brain to train for your specific problem.
 
 

    1.6   PolyNet's Version
The PolyNet has several versions:

Version                   Neurons                                  prices

_____________________________________________

Standard version      10,000                                  $129.99
100K version           100,000                                $499
250K version            250,000                               $999
1M version               1,000,000                             $9999
Other customized versions                                     $999 and up

Consulting fee:

The free customer support covers how to operate your PolyNet
If you need help on how to encode your problem, the consulting fees are:
  • $250 per problem if the data has 1000 rows or less;
  • $500 per problem if the data has 1,001-10,000 rows or less;
  • $1000 per problem if the data has more than 10,000 rows.

 
 
Online Order or if you prefer, Online Fax order: 
PolyNet for Windows 95/98 
Version 4.0 ($124.99 + $5 US Shipping and Handling)

Mail Order: $129.99 (S&H included)

PolyNet 4.0
Attrasoft, Attn.: Gina
P. O. Box 13051
Savannah, GA. 31406, USA
Products New Lottery Stock jpg/gif Order Order Help Demos
Please join Attrasoft Discussion Group.
Send questions or comments to:
webmaster@attrasoft.com
or contact us at: 
          Attrasoft, P. O. Box 13051, Savannah, GA. 31406, USA
Copyright © 1998 Attrasoft, Inc. All rights reserved.