Hardware architectures for deep learning /

This book discusses innovative ideas in the design, modelling, implementation, and optimization of hardware platforms for neural networks. The book provides an overview of this emerging field, from principles to applications, for researchers, postgraduate students and engineers who work on learning-...

Descripción completa

Detalles Bibliográficos
Clasificación:Libro Electrónico
Otros Autores: Daneshtalab, Masoud, 1979- (Editor ), Modarressi, Mehdi (Editor )
Formato: Electrónico eBook
Idioma:Inglés
Publicado: Stevenage, United Kingdom : The Institution of Engineering and Technology, 2020.
Colección:Materials, circuits and devices series ; 55.
Temas:
Computer architecture.
Machine learning.
Neural networks (Computer science)
Neural Networks, Computer
Machine Learning
Ordinateurs > Architecture.
Apprentissage automatique.
Réseaux neuronaux (Informatique)
Computer architecture
Machine learning
feedforward neural nets.
learning (artificial intelligence)
neural chips.
neural net architecture.
neuromorphic engineering.
recurrent neural nets.
Acceso en línea:Texto completo
Tabla de Contenidos:
  • Intro
  • Contents
  • About the editors
  • Preface
  • Acknowledgments
  • Part I. Deep learning and neural networks: concepts and models
  • 1. An introduction to artificial neural networks / Ahmad Kalhor
  • 1.1 Introduction
  • 1.1.1 Natural NNs
  • 1.1.2 Artificial neural networks
  • 1.1.3 Preliminary concepts in ANNs
  • 1.2 ANNs in classification and regression problems
  • 1.2.1 ANNs in classification problems
  • 1.2.2 ANNs in regression problems
  • 1.2.3 Relation between classification and regression
  • 1.3 Widely used NN models
  • 1.3.1 Simple structure networks
  • 1.3.2 Multilayer and deep NNs
  • 1.4 Convolutional neural networks
  • 1.4.1 Convolution layers
  • 1.4.2 Pooling layers
  • 1.4.3 Learning in CNNs
  • 1.4.4 CNN examples
  • 1.5 Conclusion
  • References
  • 2. Hardware acceleration for recurrent neural networks / Sima Sinaei and Masoud Daneshtalab
  • 2.1 Recurrent neural networks
  • 2.1.1 Long short-term memory
  • 2.1.2 Gated recurrent units
  • 2.2 Hardware acceleration for RNN inference
  • 2.2.1 Software implementation
  • 2.2.2 Hardware implementation
  • 2.3 Hardware implementation of LSTMs
  • 2.3.1 Model compression
  • 2.3.2 Datatype and Quantization
  • 2.3.3 Memory
  • 2.4 Conclusion
  • References
  • 3. Feedforward neural networks on massively parallel architectures / Reza Hojabr, Ahmad Khonsari, Mehdi Modarressi, and Masoud Daneshtalab
  • 3.1 Related work
  • 3.2 Preliminaries
  • 3.3 ClosNN: a customized Clos for neural network
  • 3.4 Collective communications on ClosNN
  • 3.5 ClosNN customization and area reduction
  • 3.6 Folded ClosNN
  • 3.7 Leaf switch optimization
  • 3.8 Scaling to larger NoCs
  • 3.9 Evaluation
  • 3.9.1 Performance comparison under synthetic traffic
  • 3.9.2 Performance evaluation under realistic workloads
  • 3.9.3 Power comparison
  • 3.9.4 Sensitivity to neural network size
  • 3.10 Conclusion
  • References
  • Part II. Deep learning and approximate data representation
  • 4. Stochastic-binary convolutional neural networks with deterministic bit-streams / M. Hassan Najafi, S. Rasoul Faraji, Bingzhe Li, David J. Lilja, and Kia Bazargan
  • 4.1 Overview
  • 4.2 Introduction
  • 4.3 Background
  • 4.3.1 Stochastic computing
  • 4.3.2 Deterministic low-discrepancy bit-streams
  • 4.3.3 Convolutional neural networks
  • 4.4 Related work
  • 4.5 Proposed hybrid binary-bit-stream design
  • 4.5.1 Multiplications and accumulation
  • 4.5.2 Handling negative weights
  • 4.6 Experimental results
  • 4.6.1 Performance comparison
  • 4.6.2 Cost comparison
  • 4.7 Summary
  • Acknowledgment
  • References
  • 5. Binary neural networks / Najmeh Nazari and Mostafa E. Salehi
  • 5.1 Introduction
  • 5.2 Binary neural networks
  • 5.2.1 Binary and ternary weights for neural networks
  • 5.2.2 Binarized and ternarized neural networks
  • 5.3 BNN optimization techniques
  • 5.4 Hardware implementation of BNNs
  • 5.5 Conclusion
  • References
  • Part III. Deep learning and model sparsity
  • Part IV. Convolutional neural networks for embedded systems
  • Part V. Deep learning on analog accelerators.

Ejemplares similares