2.   Attrasoft Component-Object Structures

2.1   Neural Network Layer
2.2   Presentation Layer
2.3   Application Layer

2.3.1   ABM 2.7
2.3.2   PolyNet 4.0
2.3.3   DecisionMaker 2.5
2.3.4   Predictor 2.6

2.4   What is PolyApplet Class Library?
2.5   Organization Of This User�s Guide
2.6   Terminology
2.7   Chapter Overview

---

2.   Attrasoft Component-Object Structures

Attrasoft Component-Object structure consists of three layers:

 Application Layer
 Presentation Layer
 Neural Network Layer

This software will use the following software to illustrate the Attrasoft objects structure:

ABM 2.7
DecisionMaker 2.5
Predictor 2.6
PolyNet 4.0

2.1   Neural Network Layer

There are two neural net modules:

 ABM (Attrasoft Boltzmann Machine): binary "Boltzmann Machine" and "Hopfield Model ";
 PolyNet: multi-valued "Boltzmann Machine" and "Hopfield Model ".

The Attrasoft PolyNet simulates neurons with 16-bit integer values, while the ABM simulates binary neurons.

The library is called:

 Attrasoft.PolyApplet63

And the main class is:

Attrasoft.PolyApplet63.PolyApplet63.

2.2   Presentation Layer

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.

Examples of functions in this layer are:

• Images to neural data;
• Numerical data to neural data;
• Video data to neural data;
• �

This software will focus on numerical data. The Presentation (Translation) layer class library is:

Attrasoft.PolyApplet63.abm2

And the main class is:

Attrasoft.PolyApplet63.abm2.abm2Main.

2.3   Application Layer

This is the front-end graphical user interface, which the users see directly.

2.3.1   ABM 2.7

ABM 2.7 has two layers:

 Application Layer: ABM 2.7
 Neural Network Layer: PolyApplet63 (binary neurons)

The presentation layer is absent and the user provides neural data directly to the PolyApplet63 component.

2.3.2   PolyNet 4.0

Attrasoft PolyNet has two layers:

 Application Layer: PolyNet 4.0
 Neural Network Layer: PolyApplet63 (multi-valued neurons)

The presentation layer is absent and the user provides neural data directly to the PolyApplet63 component.

2.3.3   DecisionMaker 2.5

The Attrasoft DecisionMaker has three layers:

Application Layer: DecisionMaker 2.5
Presentation Layer: Abm2
Neural Network Layer: PolyApplet63 (binary neurons)

2.3.4   Predictor 2.6

The Attrasoft Predictor has three layers:

 Application Layer: Predictor 2.6
 Presentation Layer: Abm2
 Neural Network Layer: PolyApplet63 (binary neurons)

2.4   What is PolyApplet Class Library?

The Attrasoft PolyApplet class library has three parts:

• PolyApplet63 (�binary neurons� + �multi-valued neurons�)
• Abm2
• Predictor63, DecisionMaker63, PredictorPoly63, DecisionMakerPoly63

This class library implements Attrasoft Component-Object structure, which consists of three layers:

 Application Layer
 Presentation Layer
 Neural Network Layer

PolyApplet63 is the Neural Network Layer class library. This library is further divided into two parts:

• Binary neurons
• Multi-valued neurons

Abm2 is the Presentation Layer class library.

The Application Layer class libraries are:

• Predictor63
• DecisionMaker63
• PredictorPoly63
• DecisionMakerPoly63.

2.5   Organization Of This User�s Guide

This class library is organized in three parts:

Neural Network Layer
Presentation Layer
Application Layer

Part I, Neural Network Layer, consists of chapters, 3, 4, 5, 6, and 7.
Part II, Presentation Layer, consists of chapters 8, 9, 10, and 11.
Part III, Layer, Application consists of chapters 12, 13, 14, and 15.

2.6   Terminology

Predictor

Predictor refers to the standalone software, Attrasoft Predictor For Windows, 2.6.

DecisionMaker

DecisionMaker refers to standalone software, Attrasoft DecisionMaker for Windows, 2.5.

ABM

ABM refers to standalone software, Attrasoft ABM for Windows, 2.7.

PolyNet

PolyNet refers to standalone software, Attrasoft PolyNet for Windows, 4.0.

ImageFinder

ImageFinder refers to standalone software, Attrasoft ImageFinder for Windows, 6.0.

PolyApplet

PolyApplet refers to class library, Attrasoft PolyApplet, 6.0. This class library is consists of:
• PolyApplet63 (abm + polynet)
• Abm2
• Predictor63
• DecisionMaker63
• PredictorPoly63
• DecisionMakerPoly63

PolyApplet63

PolyApplet63 refers to the Neural Layer component of the class library, Attrasoft PolyApplet, 6.0. This class library is consists two parts of:
• Binary neurons
• Multi-valued neurons

Abm2

Abm2 refers to the Presentation Layer component of the class library, Attrasoft PolyApplet, 6.0.

DecisionMaker63

DecisionMaker63 refers to the Application Layer component of the class library, Attrasoft PolyApplet, 6.0.

Predictor63

Predictor63 refers to the Application Layer component of the class library, Attrasoft PolyApplet, 6.0.

DecisionMakerPoly63

DecisionMakerPoly63 refers to the Application Layer component of the class library, Attrasoft PolyApplet, 6.0.

PredictorPoly63

PredictorPoly63 refers to the Application Layer component of the class library, Attrasoft PolyApplet, 6.0.

Abm63

The neural network layer class library is PolyApplet63. This class library has two groups of commands, with one group dealing with binary neural net and the other group dealing with multi-valued neural net. Abm63 refers to a component of the class library, PolyApplet63, which deals with the binary commands.

PolyNet63

The neural network layer class library is PolyApplet63. This class library has two groups of commands, with one group dealing with binary neural net and the other group dealing with multi-valued neural net. PolyNet63 refers to a component of the class library, PolyApplet63, which deals with the multi-valued commands.

2.7   Chapter Overview

Part I � Neural Layer:

Chapter 3 introduces the PolyApplet63, the neural network layer class library. This class is further divided into two algorithms:

• Abm63 (Binary Boltzmann Machine) and
• PolyNet63 (Multi-valued Boltzmann Machine).

Chapter 4 attempts to build software like a simple version of the Attrasoft ABM 2.7, using PolyApplet63 (binary part).

Chapter 5 attempts to build software like a simple version of the Attrasoft PolyNet 4.0, using PolyApplet63 (multi-valued part).

Chapter 6 attempts to build software like the Attrasoft PolyNet 4.0, using PolyApplet63, i.e. add more functions to chapter 5.

Chapter 7 will use the Array commands for training and recognition. Array commands provide the neural net with data through array parameters, while File commands provide the neural net with data through files.

• Array commands are used in the Attrasoft ImageFinder; and
• File commands are used in the Predictor and DecisionMaker.

Part II � Presentation Layer:

Chapter 8 introduces the class library Abm2. It also attempts to build software like the Attrasoft Predictor 2.6, using PolyApplet63 (binary algorithm) and abm2. This chapter is focused on how to translate the Predictor input data into PolyApplet63 (binary algorithm) input data and how to translate the PolyApplet63 (binary algorithm) output data to Predictor output data.

Chapter 9 attempts to build prediction software using PolyApplet63 (multi-valued algorithm) and Abm2. This chapter is focused on how to translate the Predictor input data into PolyApplet63 (multi-valued algorithm) input data and how to translate the PolyApplet63 (multi-valued algorithm) output data to Predictor output data. The difference between chapter 8 and 9 is that two different algorithms are used at the neural net layer.

Chapter 10 attempts to build software like the Attrasoft DecisionMaker 2.5, using PolyApplet63 (binary) and Abm2. This chapter is focused on how to translate the DecisionMaker input data into PolyApplet63 (binary) input data and how to translate the PolyApplet63 (binary) output data to DecisionMaker output data.

Chapter 11 attempts to build function-inference software, using PolyApplet63 (multi-valued) and Abm2. This chapter is focused on how to translate the DecisionMaker input data into PolyApplet63 (multi-valued) input data and how to translate the PolyApplet63 (multi-valued) output data to DecisionMaker output data. The difference between chapter 10 and 11 is that two different algorithms are used at the neural net layer.

Part III � Application Layer:

Chapter 12 introduces the class library �Predictor63�. It also attempts to build software like the Attrasoft Predictor 2.6, using Predictor63.

Chapter 13 introduces the class library �PredictorPoly63�. It also attempts to build prediction software similar to the Attrasoft Predictor 2.6, with binary neural net being replaced by multi-valued neural net. The difference between chapter 12 and 13 is that two different algorithms are used at the neural net layer.

Chapter 14 introduces the class library �DecisionMaker63�. It also attempts to build software like the Attrasoft DecisionMaker 2.5, using DecisionMaker63.

Chapter 15 introduces the class library �DecisionMaker63�. It also attempts to build function-inference software similar to the Attrasoft DecisionMaker 2.5, with binary neural net being replaced by multi-valued neural net. The difference between chapter 14 and 15 is that two different algorithms are used at the neural net layer.
 

Return