Skip to main content
main-content

Über dieses Buch

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 modelling in bio-neuroinformatics; multisensory streaming data modelling 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.

Inhaltsverzeichnis

Frontmatter

Time-Space and AI. Artificial Neural Networks

Frontmatter

Chapter 1. Evolving Processes in Time-Space. Deep Learning and Deep Knowledge Representation in Time-Space. Brain-Inspired AI

Abstract
This chapter presents the challenges to information sciences when dealing with complex evolving processes in time-space. The emphasis here is on processes/systems that evolve/develop/unfold/change in time-space and what characterises them. To model such processes, to extract deep knowledge that drives them and to trace how this knowledge changes over time, are among the main objectives of the brain-like approach that we take in this book by using SNN. And before going to SNN in the next chapters, we introduce how evolving processes can be represented as data, information and knowledge, and more specifically, what is deep knowledge that we will target to achieve through deep learning in SNN.
Nikola K. Kasabov

Chapter 2. Artificial Neural Networks. Evolving Connectionist Systems

Abstract
Classical artificial neural networks (ANN) were developed to learn from data. Evolving connectionist systems (ECOS) were further developed by the author and taken further by other researchers not only to learn in an adaptive, incremental way from data that measure evolving processes, but to extract rules and knowledge from the trained systems. Both methods were initially inspired by some principles of learning in the brain, but then they were developed mainly as machine learning and AI tools and techniques, with a wider scope of applications. Many of the architectures and learning methods of ANN and ECOS were used in the development of SNN, deep learning systems and brain-inspired AI discussed in other chapters of the book.
Nikola K. Kasabov

The Human Brain

Frontmatter

Chapter 3. Deep Learning and Deep Knowledge Representation in the Human Brain

Abstract
Spiking neural networks (SNN) and the deep learning algorithms for them have been inspired by the structure, the organisation and the many aspects of deep learning and deep knowledge representation in the human brain. This chapter presents basic information about brain structures and functions and reveals some inner processes of deep learning and deep knowledge representation as inspiration for brain-inspired SNN (BI-SNN) and brain-inspired AI (BI-AI) presented in the next chapters. The presented here information is not intended for modeling the brain in its precise structural and functional complexity, but rather for: (1) Borrowing spatio-temporal information processing principles from the brain for the creation of brain-inspired SNN and brain-inspired AI as general spatio-temporal data machines for deep learning and deep knowledge representation in time-space; (2) Understanding brain data, when modeled with SNN, for a more accurate analysis and for a better understanding of the brain processes that generated the data.
Nikola K. Kasabov

Spiking Neural Networks

Frontmatter

Chapter 4. Methods of Spiking Neural Networks

Abstract
Spiking neural networks (SNN) are biologically inspired ANN where information is represented as binary events (spikes), similar to the event potentials in the brain, and learning is also inspired by principles in the brain. SNN are also universal computational mechanisms (Maass in Math Found Comput Sci 1998, 72–83, 1998 [1]). These and many other reasons that are discussed in this chapter make SNN a preferred computational paradigm for modelling temporal and spatio-temporal data and for building brain-inspired AI. This chapter gives the background information for SNN that is further used in the rest of the book.
Nikola K. Kasabov

Chapter 5. Evolving Spiking Neural Networks

Abstract
Evolving SNN (eSNN) are a class of SNN and also a class of ECOS (Chap. 2) where spiking neurons are created (evolved) and merged in an incremental way to capture clusters and patterns from incoming data. This gives a new quality of the SNN systems to become adaptive, fast trained and to capture meaningful patterns from the data, departing the “curse of the black box neural networks’ and the “curse of catastrophic forgetting” as manifested by some traditional ANN models (Chap. 2). The inspiration comes from the brain as the brain always evolves its structure and functionality through continuous learning. It is always evolving and forming new knowledge.
Nikola K. Kasabov

Chapter 6. Brain-Inspired SNN for Deep Learning in Time-Space and Deep Knowledge Representation. NeuCube

Abstract
This chapter introduces brain-inspired evolving SNN (BI-SNN) in which both the SNN architecture and learning are inspired by the structure, organisation and learning in the human brain. BI-SNN manifest deep learning from data and deep knowledge representation inspired by human brain as discussed in Chap. 3 of the book. In BI-SNN data is represented as spikes, information is represented as spatio-temporal spike patterns and deep knowledge is represented as patterns of connections that are subject to deep learning and can be interpreted by humans.
Nikola K. Kasabov

Chapter 7. Evolutionary- and Quantum-Inspired Computation. Applications for SNN Optimisation

Abstract
The chapter introduces the main principles and several algorithms of both evolutionary computation (EC) and its further development as quantum inspired evolutionary computation (QiEC).
Nikola K. Kasabov

Deep Learning and Deep Knowledge Representation of Brain Data

Frontmatter

Chapter 8. Deep Learning and Deep Knowledge Representation of EEG Data

Abstract
This chapter presents general methods for deep learning and deep knowledge representation of EEG data in brain-inspired SNN (BI-SNN). These methods are applied to develop specific methods for EEG data analysis and for modelling brain cognitive functions, such as: performing cognitive tasks; emotion recognition from face expression; sub-conscious processing of stimuli; modelling attentional bias.
Nikola K. Kasabov

Chapter 9. Brain Disease Diagnosis and Prognosis Based on EEG Data

Abstract
This chapter applies the methodology for learning and pattern recognition with BI-SNN, introduced in Chap. 8 on EEG data measuring changes in brain states due to a brain disease or treatment.
Nikola K. Kasabov

Chapter 10. Deep Learning and Deep Knowledge Representation of fMRI Data

Abstract
The chapter presents first background information about functional magnetic-resonance imaging (fMRI) and then introduces methods for deep learning and deep knowledge representation from fMRI data using brain-inspired SNN. These methods are applied to develop specific methods for fMRI data analysis related to cognitive processes.
Nikola K. Kasabov

Chapter 11. Integrating Time-Space and Orientation. A Case Study on fMRI + DTI Brain Data

Abstract
This chapter introduces a new method for the integration of time-space data with additional and sometimes, a priory existing information, about the orientation (direction) of the spread of the temporal information.
Nikola K. Kasabov

SNN for Audio-Visual Data and Brain-Computer Interfaces

Frontmatter

Chapter 12. Audio- and Visual Information Processing in the Brain and Its Modelling with Evolving SNN

Abstract
This chapter presents first some background knowledge on how the human brain processes audio- and visual information. Then methods are presented for audio-, visual- and for the integrated audio and visual information processing using evolving spiking neural networks that include convolutional evolving spiking neural networks (CeSNN). Case studies are presented for person identification.
Nikola K. Kasabov

Chapter 13. Deep Learning and Modelling of Audio-, Visual-, and Multimodal Audio-Visual Data in Brain-Inspired SNN

Abstract
This chapter presents methods for audio-, visual- and for the integrated audio and visual information processing using brain-inspired SNN architectures such as NeuCube. Case studies are presented for short musical pieces recognition, fast moving object recognition, age-invariant face identification, moving digits recognition and other.
Nikola K. Kasabov

Chapter 14. Brain-Computer Interfaces Using Brain-Inspired SNN

Abstract
This chapter presents methods of BI-SNN for brain-computer interfaces (BCI). It introduces a new types of BCI, called brain-inspired BCI (BI-BCI). The BI-BCI can not only classify brain signals in a ‘black-box’ as the traditional BCI do, but they can create a model of the brain signals when a person is performing a task providing a neurofeedback, enabling a better understanding of the brain activities.
Nikola K. Kasabov

SNN in Bio- and Neuroinformatics

Frontmatter

Chapter 15. Computational Modelling and Pattern Recognition in Bioinformatics

Abstract
This chapter explores the ability of SNN to capture changes in Bioinformatics data for predicting events or classifying biological states from DNA, gene and protein data. It starts with a bioinformatics primer.
Nikola K. Kasabov

Chapter 16. Computational Neuro-genetic Modelling

Abstract
While Chap. 15 gave the basics of molecular biology and some methods for modelling bioinformatics data, computational neurogenetic modelling (CNGM) takes inspiration from neuro-genetics and develops neural network models that include gene information in their structure and functionality, similar to the biological neural networks, that have genes in the nucleus of each neuron, that not only affect but also cause the spiking activity of the neurons. CNGM is a new science direction with promising applications, some of them discussed in the chapter.
Nikola K. Kasabov

Chapter 17. A Computational Framework for Personalised Modelling. Applications in Bioinformatics

Abstract
The chapter presents a computational framework for building personalised models (PM) for accurate prediction of an outcome for the individual. First, a general scheme for building PM using integrated feature and model parameter optimisation is presented. The framework is used to develop two specific methods using: (a) traditional ANN techniques; (b) using evolving spiking neural networks (eSNN). Both methods are illustrated on benchmark biomedical data.
Nikola K. Kasabov

Chapter 18. Personalised Modelling for Integrated Static and Dynamic Data. Applications in Neuroinformatics

Abstract
The chapter presents methods for building personalised models (PM) for accurate prediction of an outcome for the individual. The general framework for PM from Chap. 17 is here further developed for using brain inspired SNN architectures (BI-SNN). The latter ones facilitate integrated modelling of both static and dynamic (temporal) data related to an individual and groups of individuals. Case studies on predicting stroke and response to treatment are presented in details.
Nikola K. Kasabov

Deep in Time-Space Learning and Deep Knowledge Representation of Multisensory Streaming Data

Frontmatter

Chapter 19. Deep Learning of Multisensory Streaming Data for Predictive Modelling with Applications in Finance, Ecology, Transport and Environment

Abstract
This chapter presents methods for using eSNN and BI-SNN for deep, incremental learning and predictive modelling of streaming data and for deep knowledge representation. The methods are applied for predictive modelling in the areas of finance, ecology, transport and environment using respective multisensory streaming data. Each of these applications require specific model design in terms of data preparation, SNN model parameters, experimental setting and validation. Each of the methods are illustrated with case study problems and data, but their applicability can be extended to a wider class of problems where multisensory streaming data is available.
Nikola K. Kasabov

Future Development in BI-SNN and BI-AI

Frontmatter

Chapter 20. From von Neumann Machines to Neuromorphic Platforms

Abstract
Spiking neural networks (SNN), being highly parallel computational systems, can be implemented on various computational platforms, from the traditional von Neumann machines to the specialised neuromorphic platforms.
Nikola K. Kasabov

Chapter 21. From Claude Shannon’s Information Entropy to Spike-Time Data Compression Theory

Abstract
This chapter of the book proposes a new information theory for temporal data compression through spike-time encoding for the purpose of reducing the amount of raw data from time series but preserving the information in terms of accuracy of pattern recognition and pattern classification. Most of the data in information sciences are temporal or spatio/spectro temporal, such as brain data, audio and video data, environmental and ecological data, financial and social data, etc. as discussed in the other chapters of the book and the proposed data compression method is applicable to all of them.
Nikola K. Kasabov

Chapter 22. From Brain-Inspired AI to a Symbiosis of Human Intelligence and Artificial Intelligence

Abstract
The chapter argues that SNN allow for the integration of all levels of information processing in the brain and in nature, from quantum, molecular and neuro-genetic, to brain signals, evolution and consciousness. The chapter presents future directions for using SNN to build brain-inspired AI systems that are able to both receive and communicate knowledge with humans for a symbiotic and collaborative work, led by the human intelligence (HI).
Nikola K. Kasabov

Backmatter

Weitere Informationen

Premium Partner

    Bildnachweise