Neural Networks

Logical Designs gives courses on site for neural network applications. See course outline below.

Data Fusion

Data Fusion training and seminars. See course outline below.

Download Data Fusion course announcement. Multi Intelligent Fusion (PDF)

Neural Network Course Outline

I. The Role for Neural Networks in DF&RM

a. Engineering Data Fusion & Resource Management (DF&RM) Solutions

1. DF&RM Role

2. DF&RM Networks

3. DF&RM Nodes

b. DF&RM Toolbox of Techniques

1. Probabilistic

2. Possibilistic

3. Expert Systems

4. Neural Networks

c. Role for Neural Networks

II. What is a Neural Network?

a. Basic Characteristics

1. The Processing Element

2. The Magic of the Transfer Function

3. Weight Adjustment Basics

b. Network Learning Types

1. Supervised Learning

2. Un-supervised Learning

3. Graded or Reinforcement Learning

c. Network Architectures

d. Comparison to Traditional Methods

III. Using Neural Networks with Data

a. From Information to Inputs

1. The Curse of Dimensionality

2. Basic Input Types for Data Representation

3. Scaling of Data

b. Generalization

1. How Much Data is Enough

2. Size of the Network

3. Variable Selection

c. Validation

IV. Sample Neural Networks Applications

a. Pattern Recognition Applications

b. Image Understanding NN’s

c. Spacecraft Structure Neurocontrol

V. Spacecraft Anomaly Detection & Recognition NN’s

a. Anomaly Detection Training & Testing

b. Anomaly Recognition Training & Testing

VI. Working with ThinksPro Software and Toolkit

a. Setting Training Parameters

b. Preparing Training Data

c. Training the Network

d. Testing the Network

VII. Training the Anomaly Detection and Recognition System

a. Building the Training/Test Sets from Data

b. Training the Anomaly Detection Net

c. Post Processing of Outputs and Reports

d. Building the Classification Net training data

e. Training the Classification net

f. Classification Reports

Data Fusion Course Outline

1. Motivation and Objectives

What is data fusion and resource management (DF&RM)?

Key DF&RM Needs and CONOPS

The Role for DF&RM Processes

Need for a DF&RM Applications Layer Software Architecture

Need for Data Fusion to Bootstrap Resource Management

Use of DNN Architecture to Develop Performance Evaluation Systems

2. Data Fusion & Resource (DF&RM) Management Architectures

DF&RM Architecture Selection Criteria

Comparison of Alternative Architectures and Models

Boyd Observe, Orient, Decide, Act (OODA) Loop Model
JDL Data Fusion Model
Fusion & Management Dual Node Network (DNN) Architecture
•Dasarathy Model
Bedworth and O'Brien's Omnibus Model
Drummond/Kovacich Taxonomy

· AFRL Information Fusion Architecture

Selection of the DNN Architecture

3. The DF&RM Dual Node Network (DNN) Architecture

Role for the DF&RM DNN Architecture Within Layered Migration Architectures

DF&RM Architecture Components & Interfaces

· The Data Fusion Tree Paradigm

o Level 0: Sub-object Fusion (emitters, pixels)

o Level 1: Object Level Fusion

o Level 2: Multi-Object Situation Assessment

o Level 3: Impact Assessment

· “Dual” Resource Management Tree

o Level 0: Resource Component Management (modes)

o Level 1: Resource Mode Management (sensors, countermeasures, process)

o Level 2: Multi-Resource Management

o Level 3: Operation Effects Management

DF&RM System Engineering Guidelines

· Phased Spiral Development & Rapid Prototyping

DF&RM Systems Engineering Guidelines

· System Role Optimization

o Concept of Operations (CONOPS)

o “Black Box” System Description

o Proficiency Feedback

· DF&RM Network Design Optimization

o Requirements Refinement

o DF&RM Network Design

o Measures of Effectiveness (MOE) Performance Feedback

· DF&RM Node Optimization

o Requirements Refinement

o DF&RM Node Design

o Measures of Performance (MOP) Feedback

· Detailed Design (Pattern) Optimization

o Requirements Refinement

o Detailed (Pattern) Design

o Refined MOP Feedback

Problem-to-Solution Space Map

Technique Decision Flow Diagrams

Expert System Design to Support DF&RM Software Development.

4. The Role for DF&RM in Applications

Need for a Consistent Tactical Picture

Army Warfighter Situation Awareness and Decision Support Needs

Intelligence Support RTIC/RTOC Needs

Multi-spectral/Multi-sensor/Multi-target Distributed DF&RM CONOPS

Air Supremacy and Precision Strike DF&RM Role

Army Distributed DF&RM Role

Ballistic Missile Defense DF&RM Role

5. DF&RM Distributed Network Optimization

Fighter-Based DF&RM (radar, IR, IFF, RWR, MWS, CNI) Networks

Intelligence, Surveillance, and Reconnaissance (ISR) Networks

Army Battlespace DF&RM Networks

Ballistic Missile Defense DF&RM Networks

6. Data Fusion Node Optimization

Data Fusion Node

· Common Referencing (data mediation, alignment)

· Data Association (hypothesis generation, evaluation, selection)

· State Estimation (kinematics, parametrics, ID, attributes)

Resource Management Node

· Task Preparation (task mediation, conflict resolution)

· Task Planning (plan hypothesis generation, evaluation, selection)

· Resource Control (continuous and discrete mode control)

Data Association Problem Space

Data Association Solution Space

· Hypothesis Generation Techniques (performance versus cost)

· Hypothesis Evaluation Techniques

o Bayesian Scoring (max a posteriori versus chi-square, noncommensurate attributes, ID pedigree, report/track confidences)

o Possibilistic Scoring (fuzzy, evidential)

o Symbolic/Logic Association (scripts, rules)

o Neural Net Pattern Recognition (inner product, pulse-stream)

· Hypothesis Selection Search (assignment, set covering, n-dim relaxation)

State Estimation Problem Space

State Estimation Solution Space

· Kinematics Track State Estimation

· Entity Type Class Tree Updating

· Track Confidence State Updating (probability of false track and coverage)

Situation Awareness Fusion Node Processing