This book intends to report the new results of the efforts on the study of Layered Intelligence of the Machine Brain (LIMB). The book collects novel research ideas in LIMB and summarizes the current machine intelligence level as “five layer intelligence”- environments sensing, active learning, cognitive computing, intelligent decision making and automatized execution. The book is likely to be of interest to university researchers, R&D engineers and graduate students in computer science and electronics who wish to learn the core principles, methods, algorithms, and applications of LIMB.
Author(s): Wen-Feng Wang, Xi Chen, Tuozhong Yao
Series: Research on Intelligent Manufacturing
Publisher: Springer
Year: 2022
Language: English
Pages: 222
City: Singapore
Preface
Contents
1 Structure of a Machine Brain
1.1 Structure of the Human Brain
1.1.1 The External Structure
1.1.2 The Psychological Structure
1.2 From Human Brain to Machine Brain
1.2.1 The Physiological Simulation
1.2.2 Learning Mechanisms of a Brain
1.2.3 To Simulate the Learning Mechanisms
1.3 Intelligence Layers—Conceptual Framework
1.3.1 The Structure of Neurons
1.3.2 Updates of the Synaptic Weights
1.3.3 Memory Mechanisms of the Brain
1.4 Outline of the Book
References
2 The First Intelligence Layer—Environments Sensing
2.1 Introduction
2.2 Design of the Sensing System
2.2.1 Description of the Functional Zones
2.2.2 The Overall Structure of the Sensing System
2.3 The Basic Principles for Environments Sensing
2.3.1 Analog-To-Digital Conversion
2.3.2 Sampling and Interpolation
2.3.3 Information Display
2.4 Implementation of Environments Sensing
2.4.1 Basic Preparations
2.4.2 Output and Display
2.4.3 The Human–Machine Interface
2.5 Concluding Remarks and a Next Research Priority
References
3 The Second Intelligence Layer—Active Learning
3.1 Introduction
3.2 The Basic Learning System
3.2.1 The System Structure
3.2.2 Connection and Sharing
3.2.3 Regression and Propagation
3.3 Implementation of Active Learning
3.3.1 The TensorFlow Framework
3.3.2 Design of the Learning Module
3.3.3 The Principle for Programing
3.3.4 Model Loading and Debugging
3.4 Human–Machine Interface Design
3.4.1 Preparing Workspace
3.4.2 The Supplementary Code
3.4.3 The Developing Environment
3.4.4 Designing Graphical Interface
3.5 Concluding Remarks and Unresolved Issues
References
4 The Third Intelligence Layer—Cognitive Computing
4.1 Introduction
4.2 Preliminary Interpretation
4.2.1 Beginning with the Simplest Task
4.2.2 Facial Features Computation
4.2.3 Real-Time Cognitive Computing
4.3 Understanding the Learners
4.3.1 Convex Base Learners
4.3.2 The Optimization Problem
4.3.3 The Embedding Models
4.4 Unsupervised Meta-Learning
4.4.1 Preliminary Analysis
4.4.2 The Basic Principle of Entropy Weight Method
4.4.3 Use Entropy Method to Calculate Weight
4.4.4 Modified Unsupervised Gradient-Based Learner
4.4.5 Construct Task for MAML and Proto-Networks
4.4.6 EW-Prototypical Networks
4.5 Extension of the Framework
4.5.1 Initialization
4.5.2 Sample Selection
4.5.3 CNN Parameter Update
4.5.4 Algorithm Comparison
4.5.5 Experiment 1: Fixed Label/Batch Size
4.5.6 Experiment 2: Variable Label Size
4.5.7 Experiment 3: Variable Batch Size
4.6 Concluded Remarks
References
5 The Fourth Intelligence Layer-Intelligent Decisions Making
5.1 Introduction
5.2 The System Description
5.2.1 The Basic Principle
5.2.2 Design of the Interface
5.3 Training of the Machine Brain
5.3.1 Algorithms Interpretation
5.3.2 Integration with Deep Learning
5.3.3 Implementation of the Algorithms
5.4 Decision and Prediction
5.4.1 Center and Scale Detection
5.4.2 From Decision to Prediction
References
6 The Fifth Intelligence Layer—Automatized Execution
6.1 Introduction
6.2 From Decision to Execution
6.2.1 The Basic Principle
6.2.2 An Intelligent Agent
6.2.3 Models for Execution
6.3 Design of the Automatic System
6.3.1 From Q-Learning to DQN
6.3.2 DQN with Meta-learning
6.4 Extension of the System
6.4.1 A Denoising Process
6.4.2 Processing of the Data
6.4.3 The Transform Method
6.4.4 Steps of Edge Detection
6.4.5 Image Fusion Algorithm
6.4.6 The Grayscale Stretching
References
7 Applications in Face Recognition Access Control Manufacturing
7.1 Introduction
7.2 Understanding the Scene
7.2.1 The 3D Visualization Scene
7.2.2 The Programming Scene
7.3 Establishing the Experimental Platform
7.3.1 Necessary Preparations
7.3.2 Installation of the Equipment
7.3.3 Computer and Mobile Workstation
7.4 The Models for 3D Video Fusion
7.4.1 Two-Dimensional Alignment
7.4.2 Three-Dimensional Alignment
7.4.3 Optimization and Scene Fusion
7.5 Implement of Occluded Face Recognition
7.5.1 Programming with Python
7.5.2 The System for Debugging
7.5.3 Set the Tolerance Rate for Errors
7.6 Recognition in the 3D Visualization Scene
7.6.1 Manual Modeling Process
7.6.2 Recognition Before Occlusion
7.6.3 Recognition After Occlusion
7.7 Prospects for Follow-Up Work
References
