Kasabov K.N. Time-Space, Spiking Neural Networks and Brain-Inspired Artificial Intelligence

New York: Springer, 2019. — 738 p. — (Springer Series on Bio- and Neurosystems, vol. 7). — ISBN: 978-3-662-57715-8.Spiking neural networks (SNN) are biologically inspired computational models that represent and process information internally as trains of spikes. This monograph book presents the classical theory and applications of SNN, including original author’s contribution to the area. The book introduces for the first time not only deep learning and deep knowledge representation in the human brain and in brain-inspired SNN, but takes that further to develop new types of AI systems, called in the book brain-inspired AI (BI-AI). BI-AI systems are illustrated on: cognitive brain data, including EEG, fMRI and DTI;
audio:visual data; brain-computer interfaces; personalized modeling in bio-neuroinformatics; multisensory streaming data modeling in finance, environment and ecology; data compression; neuromorphic hardware implementation. Future directions, such as the integration of multiple modalities, such as quantum-, molecular- and brain information processing, is presented in the last chapter. The book is a research book for postgraduate students, researchers and practitioners across wider areas, including computer and information sciences, engineering, applied mathematics, bio- and neurosciences.About the Book Content by Topics and Chapters and The Pathway of KnowledgeTime-Space and AI Artificial Neural Networks
Evolving Processes in Time-Space
What Are Evolving Processes?
Evolving Processes in Living Organisms
Spatio-temporal and Spectro-temporal Evolving Processes
Characteristics of Evolving Processes: Frequency, Energy, Probability, Entropy and Information
Light and Sound
Evolving Processes in Time-Space and Direction
From Data and Information to Knowledge
Defining Deep Knowledge in Time-Space
How Deep?
Statistical, Computational Modeling of Evolving Processes
Statistical Methods for Computational Modeling
Global, Local and Transductive (“Personalised”) Modeling []
Model Validation
Brain-Inspired AIClassical Artificial Neural Networks: SOM, MLP, CNN, RNN
Unsupervised Learning in Neural Networks Self-organising Maps (SOM)
Supervised Learning in ANN Multilayer Perceptron and the Back Propagation Algorithm
Convolutional Neural Networks (CNN)
Recurrent and LSTM ANN
Hybrid and Knowledge-Based ANN
Principles of ECOS
Evolving Self-organising Maps
Evolving MLP
Evolving Fuzzy Neural Networks EFuNN
Dynamic Evolving Neuro-fuzzy Inference Systems — DENFIS
Other ECOS Methods and Systems
Chapter Summary and Further Readings for Deeper KnowledgeThe Human Brain
Time-Space in the Brain
Learning and Memory
Neural Representation of Information
Perception in the Brain Is Always Spatio/Spectro-temporal
Deep Learning and Deep Knowledge Representation in Time-Space in the Brain
Information Coding
Molecular Basis of Information Processing
General Notions
Electroencephalogram (EEG) Data
CT and PET
fMRI
Chapter Summary and Further Readings for Deeper KnowledgeSpiking Neural Networks
Rate Versus Spike Time Information Representation
Spike Encoding Algorithms
Hodgkin-Huxley Model (HHM)
Leaky Integrate-and-Fire Model (LIFM)
Spike Response Model (SRM)
Probabilistic and Stochastic Spiking Neuron Models
Probabilistic Neurogenetic Model of a Neuron
Methods for Learning in SNN
SpikeProp
Spike-Time Dependent Plasticity (STDP)
Rank Order (RO) Learning Rule
Learning in Dynamic Synapses
Principles of Spike Pattern Association Learning The SPAN Model
Case Study Examples
Memory Capacity of SPAN
SPAN for Classification Problems
Why Use SNN?
Summary and Further Readings for Deeper KnowledgePrinciples and Methods of Evolving SNN (ESNN)
Convolutional ESNN (CeSNN)
Dynamic Evolving SNN (DeSNN)
Fuzzy Rule Extraction from ESNN
A Case Study of Fuzzy Rule Extraction from Water Tastant Sensory Data
Reservoir Architectures Liquid State Machines (LSM)
ESNN/DeSNN as Classification/Regression Systems for Reservoir Architectures
Chapter Summary and Further Readings for Deeper KnowledgeA General Architecture of a BI-SNN
The BI-SNN NeuCube as a Generic Spatio-temporal Data Machine
Mapping Input Temporal Variables into a D SNNcube Based on Graph Matching Optimisation Algorithm
Deep Unsupervised Learning in Time-Space and Deep Knowledge Representation from Temporal or Spatio/Spectro Temporal Data (TSTD)
Deep Supervised Learning in Time-Space
Deep Learning in Time-Space for Predictive Modeling in NeuCube The EPUSSS Algorithm
Event-Based Modeling External Versus Internal Time Past-, Present- and Future Time
A Design Methodology for Application Oriented Spatio-temporal Data Machines
Design Methodology for Implementing Application Oriented Spatio-temporal Data Machines as BI-AI Systems in NeuCube
Input Data Encoding
Spatial Mapping of Input Variables
Supervised Training and Classification/Regression of Dynamic Spiking Patterns of the SNNcube in a SNN Classifier
D Visualisation of the SNNcube
Optimisation of NeuCube Structure and Parameters
Case Studies of the Design and Implementation of Classification and Regression Spatio-temporal Data Machines
A Case Study on the Design a Regression/Prediction Spatio-temporal Data Machine in NeuCube
Chapter Summary and Further Readings for Deeper KnowledgeEvolutionary- and Quantum-Inspired Computation Applications for SNN Optimisation
The Origin and the Evolution of Life
Methods of Evolutionary Computation (EC)
Genetic Algorithms
Evolutionary Strategies (ES)
Particle Swarm Optimisation
Estimation of Distribution Algorithms (EDA)
Artificial Life Systems
Principles of Quantum Information Processing
Quantum Inspired Evolutionary Algorithm (QiEA)
Versatile QiEA (VQiEA)
Extension of the VQiEA to Deal with Continuous Value Variables
A Quantum-Inspired Representation of a SNN
Application of QiEA for the Optimisation of ESNN Classifier on Ecological Data
Integrative Computational Neuro Genetic Model (CNGM) Utilising Quantum-Inspired Representation
Quantum Inspired Particle Swarm Optimisation Algorithms
Quantum Inspired Particle Swarm Optimisation Algorithm (QiPSO) for the Optimisation of ESNN
Dynamic QiPSO
Application of DQiPSO for Feature Selection and Model Optimisation
Chapter Summary and Further Readings for Deeper KnowledgeDeep Learning and Deep Knowledge Representation of Brain Data
Spatio-temporal Brain Data
Brain Atlases
EEG Data
Deep Learning and Deep Knowledge Representation of EEG Data in BI-SNN
System Design
Experimental Results
Model Interpretation
General Notions
Using a NeuCube Model for Emotion Recognition
Analysis of the Connectivity in a Trained SNNcube When a Person Is Perceiving Emotional Face and When a Person Is Expressing Such Emotions
Deep Learning and Modeling of Peri-perceptual Processes in BI-SNN
The Psychology of Sub-conscious Brain Processes
Experimental Setting and EEG Data Collection
The Design of a NeuCube Model
ResultsSNN for Modeling EEG Data to Assess a Potential Progression from MCI to AD
Design of a NeuCube Model
Classification Results
SNN for Predictive Modeling of Response to Treatment Using EEG Data
The Case Study Problem Specification and Data Collection
Modeling the EEG Data in a NeuCube Model
Comparative Analysis of Brain Activities of MMT Subjects Under Different Drug Doses Versus CO and OP Subjects Modeling and Understanding the Information Exchange Between Brain Areas Measured Through EEG Channels
Analysis of Classification Results
Chapter Summary and Further Readings for Deeper KnowledgeWhat Are fMRI Data?
Traditional Methods for fMRI Data Analysis
Selecting Features from FMRI Data
Why Use SNN for Modeling of fMRI Spatio-temporal Brain Data?
A Methodology for Deep Learning and Deep Knowledge Representation of fMRI Data in BI-SNN
The STAR/PLUS Benchmark fMRI Data
fMRI Data Encoding, Mapping and Learning in a NeuCube SNN Model
Classification of the fMRI Data in a NeuCube-Based Model
Algorithms for Modeling fMRI Data that Measure Cognitive Processes
Connectivity Initialization and Deep Learning in a SNN Cube
A Case Study Implementation on the STAR/PLUS Data
Chapter Summary and Further Readings for Deeper KnowledgeIntroduction and Background Work
A Personalised Modeling Architecture for fMRI and DTI Data Integration Based on the NeuCube BI-SNN
Orientation-Influence Driven STDP (oiSTDP) Learning in SNN for the Integration of Time-Space and Direction, Illustrated on fMRI and DTI Data
Neuron Model
Unsupervised Weight Adaptation of Synapses
Experimental Results
Problem Specification and Data Preparation
Modeling and Experimental Results
Chapter Summary and Further Readings for Deeper KnowledgeSNN for Audio-Visual Data and Brain-Computer Interfaces
Audio and Visual Information Processing in the Human Brain
Audio Information Processing
Visual Information Processing
Integrated Audio and Visual Information Processing
Issues with Modeling Audio-Visual Information with SNN
Convolutional eSNN (CeSNN) for Modeling Visual Information
Convolutional eSNN (CeSNN) for Modeling Audio Information
Convolutional eSNN (CeSNN) for Integrated Audio-Visual Information Processing
Data Sets
Experimental Results
Chapter Summary and Further Readings for Deeper KnowledgeDeep Learning of Audio Data in the Brain
Deep Learning and Recognition of Music
Experimental Results
Two Approaches to Visual Information Processing
Applications for Fast Moving Object Recognition
Applications for Gender and Age Group Classification Based on Face Recognition
The Brain-Inspired SNN and the Proposed Retinotopic Mapping
Unsupervised and Supervised Learning of Dynamic Visual Patterns
Design of an Experiment for the MNIST-DVS Benchmark Dataset
Experimental Results
Model Interpretation for a Better Understanding of the Processes Inside the Visual Cortex
Chapter Summary and Further Readings for Deeper KnowledgeGeneral Notions
Types and Applications of BCI
The NeuCube BI-SNN Architecture
A Brain-Inspired Framework for BCI (BI-BCI) with NeurofeedbackDesign of an Experimental BI-BCI System
Analysis of the Results
General Notions
Applications
From BI-BCI to Knowledge Transfer Between Humans and MachinesSNN in Bio- and Neuroinformatics
General Notions
DNA, RNA and Proteins The Central Dogma of Molecular Biology and the Evolution of Life
Phylogenetics
The Challenges of Molecular Data Analysis
Biological Databases
General Information About Bioinformatics Data Modeling
Gene Expression Data Modeling and Profiling
Clustering of Time Series Gene Expression Data
Protein Data Modeling and Structure Prediction
General Notions
Gene Regulatory Network Modeling
Protein Interaction Networks
General Notions
A SNN Based Methodology for Gene Expression Time Series Data Modeling and Extracting GRN
Extracting GRN from a Trained Model
A Case Study Experimental Modeling of Gene Expression Time Series Data
Extracting GRN Form a Trained Model and Analysis of the GRN for New Knowledge Discovery
Discussions on the Method
Chapter Summary and Further ReadingsGeneral Notions
The PNGM of a Spiking Neuron
Using the PNGM of a Neuron to Build SNN
CNGM Architectures
The NeuCube Architecture as a CNGM
Modeling Brain Diseases
CNGM for Cognitive Robotics and Emotional Computing
Life, Death and CNGM
Chapter Summary and Further Readings for Deeper KnowledgeIntroduction: Global, Local and Personalise Modeling
A Framework for Personalised Modeling (PM) Based on Integrated Feature and Model Parameter Optimisation
Classification Accuracy and Comparative AnalysisUsing SNN and ESNN for PM
An ESNN Method for PM on Biomedical Data
A Case Study of PM for Chronic Kidney Disease Data ClassificationA NeuCube-Based Framework for PM of Integrated Static and Dynamic Data
Comparative Analysis of the NeuCube Based Method with Other Methods for PM
The Case Study Data for Individual Stroke Risk Prediction
Personalised Deep Learning and Knowledge Representation in NeuCube on the Case of Stroke
The Case Study Problem and Data
The NeuCube Based PM Model
Experimental Results
DiscussionsDeep in Time-Space Learning and Deep Knowledge Representation of Multisensory Streaming Data
A General Framework for Deep Learning and Predictive Modeling of Multisensory Time-Space Streaming Data with SNN
The Challenges of Pattern Recognition and Modeling of Multisensory Streaming Data
Modeling Streaming Data in Evolving SNN (eSNN)
A General Methodology for Modeling Multisensory Streaming Data in Brain-Inspired SNN for Classification and Regression
Stock Market Movement Prediction Using On-Line Predictive Modeling with eSNN
Early Event Prediction in Ecology: General Notions
A Case Study on Predicting Abundance of Fruit Flies Using Spatio-temporal Climate Data
NeuCube Model Creation and Modeling Results
Predictive Modeling of Seismic Data for Earthquake Forecasting Using NeuCube
Experiment Design
Discussions
Modeling Multisensory Air Pollution Streaming Data
Chapter Summary and Further Readings for Deeper KnowledgeFuture Development in BI-SNN and BI-AI
General Notions
The von Neumann Computation Principle and the Atanassov’s ABC Machine
General Principles
Hardware Platforms for Neuromorphic Computation
A Brief Overview of SNN Development Systems
The NeuCube Development System for Spatio-temporal Data Machines
Implementation of NeuCube-Based Spatio-temporal Data Machines on Traditional and on Neuromorphic Hardware Platforms
Chapter Summary and Further ReadingsClaude Shannon’s Classical Information Theory
The Proposed Information Theory for Temporal Data Compression for Classification Tasks Based on Spike-Time Encoding
A Spike-Time Encoding and Compression Method for fMRI Spatio-Temporal Data Classification
Chapter Summary and Further ReadingsTowards Integrated Quantum-Molecular-Neurogenetic-Brain-Inspired Models
Quantum Computation
The Concept of an Integrated Quantum-Neurogenetic-Brain-Inspired Model Based on SNN
Towards a Symbiosis Between Human Intelligence and Artificial Intelligence (HI + AI), Led by HI
Summary and Further Readings for a Deeper Knowledge
Epilogue