For machine intelligence applications to work successfully, machines must perform reliably under variations of data and must be able to keep up with data streams. Internet-Scale Pattern Recognition: New Techniques for Voluminous Data Sets and Data Clouds unveils computational models that address performance and scalability to achieve higher levels of reliability. It explores different ways of implementing pattern recognition using machine intelligence.
Based on the authors’ research from the past 10 years, the text draws on concepts from pattern recognition, parallel processing, distributed systems, and data networks. It describes fundamental research on the scalability and performance of pattern recognition, addressing issues with existing pattern recognition schemes for Internet-scale data deployment. The authors review numerous approaches and introduce possible solutions to the scalability problem.
By presenting the concise body of knowledge required for reliable and scalable pattern recognition, this book shortens the learning curve and gives you valuable insight to make further innovations. It offers an extendable template for Internet-scale pattern recognition applications as well as guidance on the programming of large networks of devices.
I Recognition: A New Perspective
Introduction
As We See, We Learn
Recognition at a Large Scale
Computational Intelligence Approach for Pattern Recognition
Scalability in Pattern Recognition
Distributed Approach for Pattern Recognition
Scalability of Neural Network Approaches
Key Components of DPR
System Approaches
Pattern Distribution Techniques
Current DPR Schemes
Resource Considerations for DPR Implementations
II Evolution of Internet-Scale Recognition
One-Shot Learning Considerations
One-Shot Learning Graph Neuron (GN) Scheme
One-Shot Learning Model
GN Complexity Estimation
Graph Neuron Limitations
Significance of One-Shot Learning
Hierarchical Model for Pattern Recognition
Evolution of One-Shot Learning: The Hierarchical Approach
Complexity and Scalability of A Hierarchical DPR Scheme
Reducing Hierarchical Complexity: A Distributed Approach
Design Evaluation for Distributed DPR Approach
Recognition via a Divide-and-Distribute Approach
Divide-and-Distribute Approach for One-Shot Learning IS-PR Scheme
Dimensionality Reduction in Pattern Pre-Processing
Remarks on DHGN DPR Scheme
III Systems and Tools
Internet-Scale Applications Development
Distributed Computing Models for IS-PR
Parallel Programming Techniques
From Coding to Applications
IV Implementations and Applications
Multi-Feature Classifications for Complex Data
Data Features for Pattern Recognition
Distributed Multi-Feature Recognition
Handwritten Object Classification with Multiple Features
Distributed Multi-Feature Recognition Perspective
Pattern Recognition within Coarse-Grained Networks
Network Granularity Considerations
Face Recognition using the Multi-Feature DPR Approach
Distributed Data Management within Cloud Computing
Adaptive Recognition: A Different Perspective
Event Detection within Fine-Grained Networks
Distributed Event Detection Scheme for Wireless Sensor Networks
Integrated Grid-Sensor Scheme for Structural Analysis
Distributed Event Detection: A Lightweight Approach
Recognition: The Future and Beyond
Medium of Change
Future of Internet-Scale PR
Making a Case
Bibliography
Overview
For machine intelligence applications to work successfully, machines must perform reliably under variations of data and must be able to keep up with data streams. Internet-Scale Pattern Recognition: New Techniques for Voluminous Data Sets and Data Clouds unveils computational models that address performance and scalability to achieve higher levels of reliability. It explores different ways of implementing pattern recognition using machine intelligence.
Based on the authors’ ...