SpiNNaker - a Spiking Neural Network Architecture

• Record Nr.: UNINA9910476755903321
• Autore: Furber Steve
• Pubbl/distr/stampa: Norwell, MA : , : Now Publishers, , 2020; ©2020
• Edizione: [1st ed.]
• Descrizione fisica: 1 electronic resource (350 p.)
• Collana: NowOpen
• Altri autori (Persone): BogdanPetruț
• Soggetti: Artificial intelligence
• Lingua di pubblicazione: Inglese
• Formato: Materiale a stampa
• Livello bibliografico: Monografia
• Nota di contenuto: Intro -- Copyright -- Table of Contents -- Preface -- Acknowledgements -- Funding Acknowledgements -- Glossary -- 1 Origins -- 1.1 From Ada to Alan - Early Thoughts on Brains and Computers -- 1.1.1 Ada Lovelace -- 1.1.2 Alan Turing -- 1.2 Reinventing Neural Networks - Early Thoughts on the Machine -- 1.2.1 Mighty ARMs from Little Acorns Grow -- 1.2.2 Realising Our Potential -- 1.2.3 Reinventing Neural Networks -- 1.3 The Architecture Comes Together -- 1.3.1 The State of the Neuromorphic Art -- 1.3.2 What Could We Bring to Neuromorphics? -- 1.3.3 Multicast Packet-switched AER -- 1.3.4 Optimise, Optimise… -- 1.3.5 Flexibility to Cope with Uncertainty -- 1.3.6 Big Memories -- 1.3.7 Ready to Go -- 1.4 A Scalable Hardware Architecture for Neural Simulation -- 1.4.1 Introduction -- 1.4.2 Intellectual Property -- 1.4.3 Market Opportunity -- 1.4.4 System Organisation -- 1.4.5 Node Organisation -- 1.4.6 System Architecture Issues -- 1.4.7 Development Plan -- 1.5 Summary -- 2 The SpiNNaker Chip -- 2.1 Introduction -- 2.2 Architecture -- 2.2.1 An Overview -- 2.2.2 Processor Subsystem -- 2.2.3 Router -- 2.2.4 Interconnection Networks -- 2.2.5 The Rest of the Chip -- 2.3 Multiprocessor Support -- 2.4 Event-Driven Operation -- 2.5 Chip I/O -- 2.6 Monitoring -- 2.7 Chip Details -- 2.8 Design Critique -- 2.9 Summary -- 3 Building SpiNNaker Machines -- 3.1 Putting Chips Together -- 3.1.1 SpiNN-3: Development Platform -- 3.1.2 SpiNN-5: Production Board -- 3.1.3 Nobody is Perfect: Testing and Blacklisting -- 3.2 Putting Boards Together -- 3.2.1 SpiNNaker Topology -- 3.2.2 ; spiNNlink: High-speed Serial Board-to-Board Interconnect -- 3.3 Putting Everything Together -- 3.3.1 SpiNNaker1M Assembly -- 3.3.2 SpiNNaker1M Interconnect -- 3.3.3 SpiNNaker1M Cabling -- 3.4 Using the Million-Core Machine: Tear it to Pieces -- 3.5 SpiNNaker1M in Action.; 4 Stacks of Software Stacks -- 4.1 Introduction -- 4.2 Making Use of the SpiNNaker Architecture -- 4.3 SpiNNaker Core Software -- 4.4 Booting a Million Core Machine -- 4.5 Previous Software Versions -- 4.6 Data Structures -- 4.6.1 SpiNNaker Machines -- 4.6.2 Graphs -- 4.7 The SpiNNTools Tool Chain -- 4.7.1 Setup -- 4.7.2 Graph Creation -- 4.7.3 Graph Execution -- Machine Discovery -- Mapping -- Data Generation -- Loading -- Running -- 4.7.4 Return of Control/Extraction of Results -- 4.7.5 Resuming/Running Again -- 4.7.6 Closing -- 4.7.7 Algorithms and Execution -- 4.7.8 Data Recording and Extraction -- 4.7.9 Live Interaction -- 4.7.10 Dropped Packet Re-Injection -- 4.7.11 Network Traffic Visualisation -- 4.7.12 Performance and Power Measurements -- 4.8 Non-Neural Use Case: Conway's Game of Life -- 4.9 sPyNNaker - Software for Modelling Spiking Neural Networks -- 4.9.1 PyNN -- 4.9.2 sPyNNaker Implementation -- 4.9.3 Preprocessing -- 4.9.4 SpiNNaker Runtime Execution -- Using the Low-Level Libraries -- Time-Driven Neuron Update -- Receiving a Spike -- Activation of the Spike Processing Pipeline -- Synapse processing -- Callback Interaction -- 4.9.5 Neural Modelling -- Software Structure -- Leaky Integrate and Fire Neuron -- Izhikevich Neuron -- 4.9.6 Auxiliary Application Code -- Spike Input Generation -- Simulating Extended Synaptic Delays -- 4.10 Software Engineering for Future Systems -- 4.11 Full Example Code Listing -- 5 Applications - Doing Stuff on the Machine -- 5.1 Robot Art Project -- 5.1.1 Building Brains with Nengo and Some Bits and Pieces -- 5.2 Computer Vision with Spiking Neurons -- 5.2.1 Feature Extraction -- Gabor-like Detection -- Blob Detector -- Motion Detection -- 5.3 SpiNNak-Ear - On-line Sound Processing -- 5.3.1 Motivation for a Neuromorphic Implementation -- 5.3.2 The Early Auditory Pathway.; 5.3.3 Model Algorithm and Distribution -- 5.3.4 Results -- 5.3.5 Future Developments -- 5.4 Basal Ganglia Circuit Abstraction -- 5.5 Constraint Satisfaction -- 5.5.1 Defining the Problem -- 5.5.2 Results -- 5.5.3 Graph Colouring -- 5.5.4 Latin Squares -- 5.5.5 Ising Spin Systems -- 6 From Activations to Spikes -- 6.1 Classical Models -- 6.2 Symbol Card Recognition System with Spiking ConvNets -- 6.2.1 Spiking ConvNet on SpiNNaker -- 6.2.2 Results -- 6.3 Handwritten Digit Recognition with Spiking DBNs -- Spiking DBN on SpiNNaker -- Porting DBN onto SpiNNaker -- Simulating Input Sensory Noise -- Limited Weight Precision -- 6.3.1 Results -- 6.4 Spiking Deep Neural Networks -- 6.4.1 Related Work -- 6.4.2 Siegert: Modelling the Response Function -- Biological Background -- Mismatch of the Siegert Function to Practice -- Noisy Softplus (NSP) -- 6.4.3 Generalised Off-line SNN Training -- Mapping NSP to Concrete Physical Units -- Parametric Activation Functions (PAFs) -- Training Method -- Fine Tuning -- 6.4.4 Results -- Experiment Description -- Individual Neuronal Activity -- Learning Performance -- Recognition Performance -- Power Consumption -- 6.4.5 Summary -- 7 Learning in Neural Networks -- 7.1 Sizing Up the (Biological) Competition -- 7.2 Spike-Timing-Dependent Plasticity -- 7.2.1 Experimental Evidence for Spike-Timing-Dependent Plasticity -- 7.2.2 Related Work -- 7.2.3 Implementation -- 7.2.4 Inhibitory Plasticity in Cortical Networks -- 7.2.5 The Effect of Weight Dependencies -- 7.3 Voltage-Dependent Weight Update -- 7.3.1 Results -- 7.4 Neuromodulated STDP -- 7.4.1 Eligibility Traces/Synapse Tagging -- 7.4.2 Credit Assignment -- 7.5 ; Structural Plasticity -- 7.5.1 Topographic Map Formation -- 7.5.2 Stable Mappings Arise from Lateral Inhibition -- 7.5.3 MNIST Classification in the Absence of Weight Changes.; 7.5.4 Visualisation, Visualisation, Visualisation -- 7.5.5 Rewiring for Motion Detection -- 7.6 Neuroevolution -- 7.6.1 Pac-Man on SpiNNaker -- 7.6.2 Further Exploration of NEAT -- 7.6.3 An Evolutionary Optimisation Framework for SpiNNaker -- 7.6.4 Methods -- 7.6.5 Results -- 7.6.6 Future Work -- Different EAs -- Machine Learning -- Learning-to-Learn -- Impact on Computational Neuroscience -- 8 Creating the Future -- 8.1 Survey of Currently Available Accelerators -- 8.2 SpiNNaker2 -- 8.2.1 Lessons from SpiNNaker1 -- Strengths -- Weaknesses -- 8.2.2 Scaling Performance and Efficiency -- 8.3 SpiNNaker2 Chip Architecture -- 8.4 SpiNNaker2 Packet Router -- 8.5 The Processing Element (PE) -- 8.5.1 PE Components -- Communications Controller -- Random Number Accelerator -- Rounding Accelerator -- Elementary Function (exp, log) Accelerator -- Machine Learning (ML) Accelerator -- 8.5.2 PE Implementation Strategy and Power Management -- 8.6 Summary -- References -- Index -- About the Editors -- Contributing Authors.
• Sommario/riassunto: 20 years in conception and 15 in construction, the SpiNNaker project has delivered the world’s largest neuromorphic computing platform incorporating over a million ARM mobile phone processors and capable of modelling spiking neural networks of the scale of a mouse brain in biological real time. This machine, hosted at the University of Manchester in the UK, is freely available under the auspices of the EU Flagship Human Brain Project. This book tells the story of the origins of the machine, its development and its deployment, and the immense software development effort that has gone into making it openly available and accessible to researchers and students the world over. It also presents exemplar applications from ‘Talk’, a SpiNNaker-controlled robotic exhibit at the Manchester Art Gallery as part of ‘The Imitation Game’, a set of works commissioned in 2016 in honour of Alan Turing, through to a way to solve hard computing problems using stochastic neural networks. The book concludes with a look to the future, and the SpiNNaker-2 machine which is yet to come.

The Political Economy of Reforms and the Remaking of the Proletarian Class in China, 1980s–2010s : Demystifying China's Society and Social Classes in the Post-Mao Era / / by Shan Shanne Huang

• Record Nr.: UNINA9910645890103321
• Autore: Huang Shan
• Pubbl/distr/stampa: Cham : , : Springer International Publishing : , : Imprint : Palgrave Macmillan, , 2023
• ISBN: 9783031204555; 3031204557
• Edizione: [1st ed. 2023.]
• Descrizione fisica: 1 online resource (284 pages)
• Collana: Palgrave Studies in Economic History, , 2662-6500
• Disciplina: 322.20951; 338.951009045
• Soggetti: Economic history; China - History; Economics; Marxian school of sociology; Economic History; History of China; Political Economy and Economic Systems; Marxist Sociology
• Lingua di pubblicazione: Inglese
• Formato: Materiale a stampa
• Livello bibliografico: Monografia
• Nota di contenuto: 1. Introduction -- 2. Theories and Methodology Applied -- 3. The Case of E Group Corporation – An SOE in Sichuan, post-1949 -- 4. Workers’ Returning to a Proletarian Position in post-1978 -- 5. Nostalgia” and “Protests”: Class Consciousness and Class-for-itself -- 6. Reconstruction of Classes and Class Society in China 7. Final Conclusions -- 8. Appendices.
• Sommario/riassunto: This path-breaking book unveils the true colour of China’s dominant socio-economic structure today. The author’s unique case study convincingly demonstrates the propeller behind China’s recent ‘miracle growth’. With this book, a new line of investigation can be expected to better understand post-Mao China. - Professor Kent Deng, London School of Economics, UK Shan Huang's study uses unique, in depth ; field research of the lives of workers in a state enterprise and their perception of their changed economic and political status over the era of the economic reforms since the 1980s. This work is based on intimate engagement with a specific case study, offering new insights into the development of modern China. - Professor Kerry Brown, King’s College London, UK This book comprehensively investigates the position of China’s working class between the 1980s and 2010s. It argues the case that, far from the illusion during the Maoist period that a new society had been established where the working classes held greater political and economic autonomy, economic reforms in the post-Mao era have led to the return of traditional Marxist proletariats in China. The book demonstrates how the reforms of Deng Xiaoping have led to increased economic efficiency at the expense of economic equality through an extensive case study of an SOE (state-owned enterprise) in Sichuan Province as well as wider discussions of the emergence of state capitalism on both a micro and macroeconomic level. The book also discusses workers’ protests during these periods of economic reform to reflect the reformation of class consciousness in post-Mao China, drawing on Marx’s concept of a transition from a ‘class-in-itself' to a ‘class-for-itself’. Shan Huang is a Fellow at the United Nations Development Programme in New York and a PhD candidate at King's College London, focusing on the political economy of China and Chinese economic and social history. .