Asynchronous Spiking Neurons, the Natural Key to Exploit Temporal Sparsity
| Autor: | Bernabe Linares-Barranco, Pieter Simoens, Teresa Serrano-Gotarredona, Mina A. Khoei, Bart Dhoedt, Priscila Holanda, Orlando Miguel Pires Dos Reis Moreira, Jonathan Tapson, Sam Leroux, Amirreza Yousefzadeh, Sahar Hosseini |
|---|---|
| Přispěvatelé: | Universidad de Sevilla. Departamento de Arquitectura y Tecnología de Computadores, Universidad de Sevilla. TIC178: Diseño y Test de Circuitos Integrados de Señal Mixta, European Union (UE), Ministerio de Economía y Competitividad (MINECO). España |
| Rok vydání: | 2019 |
| Předmět: |
0209 industrial biotechnology
Technology and Engineering Computer science Neural Network Inference 02 engineering and technology Asynchronous Inference Deep Neural Network Bio-inspired processing 020901 industrial engineering & automation Spiking neural network 0202 electrical engineering electronic engineering information engineering Feature (machine learning) Algorithm design and analysis Inference algorithms Electrical and Electronic Engineering Inference engine BRAIN Signal processing Artificial neural network Artificial neural networks Deep Neuromorphic hardware Convolutional Neural Network (CNN) NETWORKS Convolutional Neural Networks (CNN) Computer engineering Temporal sparsity Asynchronous communication Spiking 020201 artificial intelligence & image processing Algorithm design Neural Network (SNN) Neural networks |
| Zdroj: | idUS. Depósito de Investigación de la Universidad de Sevilla instname idUS: Depósito de Investigación de la Universidad de Sevilla Universidad de Sevilla (US) IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS |
| ISSN: | 2156-3357 2156-3365 |
| Popis: | Inference of Deep Neural Networks for stream signal (Video/Audio) processing in edge devices is still challenging. Unlike the most state of the art inference engines which are efficient for static signals, our brain is optimized for real-time dynamic signal processing. We believe one important feature of the brain (asynchronous state-full processing) is the key to its excellence in this domain. In this work, we show how asynchronous processing with state-full neurons allows exploitation of the existing sparsity in natural signals. This paper explains three different types of sparsity and proposes an inference algorithm which exploits all types of sparsities in the execution of already trained networks. Our experiments in three different applications (Handwritten digit recognition, Autonomous Steering and Hand-Gesture recognition) show that this model of inference reduces the number of required operations for sparse input data by a factor of one to two orders of magnitudes. Additionally, due to fully asynchronous processing this type of inference can be run on fully distributed and scalable neuromorphic hardware platforms. European Union's Horizon 2020 No 687299 NeuRAM European Union's Horizon 2020 No 824164 HERMES Ministerio de Economía y Competitividad TEC2015-63884-C2-1-P |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|