Event-Based Neuromorphic Systems

Neuromorphic electronic engineering takes its inspiration from the functioning of nervous systems to build more power efficient electronic sensors and processors. Event-based neuromorphic systems are inspired by the brain's efficient data-driven communication design, which is key to its quick responses and remarkable capabilities. This cross-disciplinary text establishes how circuit building blocks are combined in architectures to construct complete systems. These include vision and auditory sensors as well as neuronal processing and learning circuits that implement models of nervous systems.

Techniques for building multi-chip scalable systems are considered throughout the book, including methods for dealing with transistor mismatch, extensive discussions of communication and interfacing, and making systems that operate in the real world. The book also provides historical context that helps relate the architectures and circuits to each other and that guides readers to the extensive literature. Chapters are written by founding experts and have been extensively edited for overall coherence.

This pioneering text is an indispensable resource for practicing neuromorphic electronic engineers, advanced electrical engineering and computer science students and researchers interested in neuromorphic systems.

Key features:

• Summarises the latest design approaches, applications, and future challenges in the field of neuromorphic engineering.

• Presents examples of practical applications of neuromorphic design principles.

• Covers address-event communication, retinas, cochleas, locomotion, learning theory, neurons, synapses, floating gate circuits, hardware and software infrastructure, algorithms, and future challenges.

Auteur

Shih-Chii Liu is a group leader at the Institute of
Neuroinformatics, University of Zurich and ETH Zurich. She
received her Ph.D. in the Computation and Neural Systems program at
Caltech. She has been an instructor and topic organizer at the NSF
Telluride Neuromorphic Cognition Engineering Workshop in Telluride,
Colorado since 1998. She has also co-authored a book on analog VLSI
circuits (published by MIT Press), is an IEEE Senior member and has
held offices in a number of scientific and IEEE engineering
international conferences. Dr Liu has been working on event-based
vision and auditory sensors, multi-neuron networks, and
asynchronous circuits for more than 20 years.

Tobi Delbruck has been Professor of Physics and
Electrical Engineering at the Institute of Neuroinformatics since

1998. He leads the Sensors group which focuses on neuromorphic
      sensors and processing. He received his Ph.D. in the Computation
      and Neural Systems program at Caltech. He worked on
      electronic imaging at Arithmos, Synaptics, National Semiconductor,
      and Foveon. He co-organized the Telluride Neuromorphic Cognition
      Engineering summer workshop and the live demonstration sessions at
      ISCAS and NIPS, and is former chair of the IEEE CAS Sensory Systems
      Technical Committee. He has been awarded 9 IEEE awards and is an
      IEEE Fellow.

Giacomo Indiveri is a Professor at the University of
Zurich's Faculty of Science. He obtained his M.Sc. degree in
Electrical Engineering and his Ph.D. degree in Computer Science
from the University of Genoa, Italy. He is an ERC fellow and an
IEEE Senior member. His research interests lie in the study of real
and artificial neural processing systems, and in the hardware
implementation of neuromorphic cognitive systems, using full custom
analog and digital VLSI technology.

Adrian M. Whatley gained a degree in Chemistry at the
University of Bristol in England in 1986. After working for 10
years in the British computer industry, he took up his current
software engineering position at the Institute of Neuroinformatics
where he works primarily on asynchronous Address-Event
communication systems.

Rodney Douglas is a co-founder of the Institute of
Neuroinformatics. His central research interest over the past 25
years has been the nature of computation by the circuits of the
neocortex and their implementation both in software simulation, in
custom electronic hardware. The experimental aspect of his work has
inspired a number of cortical models of processing that use
recurrently connected neuronal architectures. He is currently
exploring principles of self-assembly in simple organisms and
circuits which he considers crucial for building truly autonomous
neuromorphic cognitive systems.

Contenu
List of Contributors xv Foreword xvii

Acknowledgments xix

List of Abbreviations and Acronyms xxi

1 Introduction 1

1.1 Origins and Historical Context 3

1.2 Building Useful Neuromorphic Systems 5

References 5

Part I UNDERSTANDING NEUROMORPHIC SYSTEMS 7

2 Communication 9

2.1 Introduction 9

2.2 Address-Event Representation 12

2.2.1 AER Encoders 13

2.2.2 Arbitration Mechanisms 13

2.2.3 Encoding Mechanisms 17

2.2.4 Multiple AER Endpoints 19

2.2.5 Address Mapping 19

2.2.6 Routing 19

2.3 Considerations for AER Link Design 20

2.3.1 Trade-off: Dynamic or Static Allocation 21

2.3.2 Trade-off: Arbitered Access or Collisions? 23

2.3.3 Trade-off: Queueing versus Dropping Spikes 24

2.3.4 Predicting Throughput Requirements 25

2.3.5 Design Trade-offs 27

2.4 The Evolution of AER Links 28

2.4.1 Single Sender, Single Receiver 28

2.4.2 Multiple Senders, Multiple Receivers 30

2.4.3 Parallel Signal Protocol 31

2.4.4 Word-Serial Addressing 32

2.4.5 Serial Differential Signaling 33

2.5 Discussion 34

References 35

3 Silicon Retinas 37

3.1 Introduction 37

3.2 Biological Retinas 38

3.3 Silicon Retinas with Serial Analog Output 39

3.4 Asynchronous Event-Based Pixel Output Versus Synchronous Frames 40

3.5 AER Retinas 40

3.5.1 Dynamic Vision Sensor 41

3.5.2 Asynchronous Time-Based Image Sensor 46

3.5.3 Asynchronous ParvoMagno Retina Model 46

3.5.4 Event-Based Intensity-Coding Imagers (Octopus and TTFS) 48

3.5.5 Spatial Contrast and Orientation Vision Sensor (VISe) 50

3.6 Silicon Retina Pixels 54

3.6.1 DVS Pixel 54

3.6.2 ATIS Pixel 56

3.6.3 VISe Pixel 58

3.6.4 Octopus Pixel 59

3.7 New Specifications for Silicon Retinas 60

3.7.1 DVS Response Uniformity 60

3.7.2 DVS Background Activity 62

3.7.3 DVS Dynamic Range 62

3.7.4 DVS Latency and Jitter 63

3.8 Discussion 64

References 67

4 Silicon Cochleas 71

4.1 Introduction 72

4.2 Cochlea Architectures 75

4.2.1 Cascaded 1D 76

4.2.2 Basic 1D Silicon Cochlea 77

4.2.3 2D Architecture 78

4.2.4 The Resistive (Conductive) Network 79

4.2.5 The BM Resonators 80

4.2.6 The 2D Silicon Cochlea Model 80

4.2.7 Adding the Active Nonlinear Behavior of the OHCs 82

4.3 Spike-Based Cochleas 83

4.3.1 Q-control of AEREAR2 Filters 85

4.3.2 Applications: Spike-Based Auditory Processing 86

4.4 Tree Diagram 87

4.5 Discussion 87

References 89

5 Locomotion Motor Control 91

5.1 Introduction 92

5.1.1 Determining Functional Biological Elements 92

5.1.2 Rhythmic Motor Patterns 93

5.2 Modeling Neural Circuits in Locomotor Control 95

5.2.1 Describing Locomotor Behavior 96

5.2.2 Fictive Analysis 97

5.2.3 Connection Models 99

5.2.4 Basic CPG Construction 100

5.2.5 Neuromorphic Architectures 102

5.3 Neuromorphic CPGs at Work 108

5.3.1 A Neuroprosthesis: Control of Locomotion in Vivo 109

5.3.2 Walking Robots 111

5.3.3 Modeling Intersegmental Coordination 112

5.4 Discussion 113

References 115

6 Learning in Neuromorphic Systems 119

6.1 Introduction: Synaptic Connections, Memory, and Learning 120

6.2 Retaining Memories in Neuromorphic Hardware 121

6.2.1 The Problem of Memory Maintenance: Intuition 121

6.2.2 The…