Modular Spiking Neural Circuits for Mapping Long Short-Term Memory on a Neurosynaptic Processor
Abstract
Due to the distributed and asynchronous nature of neural computation through low-energy spikes, brain-inspired hardware systems offer high energy efficiency and massive parallelism. One such platform is the IBM TrueNorth neurosynaptic system. Recently, TrueNorth compatible representation learning algorithms have emerged, achieving close to the state-of-the-art performance in various data sets. However, its application in temporal sequence processing models, such as recurrent neural networks (RNNs), is still only at the proof of concept level. There is an inherent difficulty in capturing temporal dynamics of an RNN using spiking neurons, which is only exasperated by the hardware constraints in connectivity and synaptic weight resolution. This paper presents a design flow that overcomes these difficulties and maps a special case of recurrent networks called long short-term memory (LSTM) onto a spike-based platform. The framework utilizes various approximation techniques, such as activation discretization, weight quantization, and scaling and rounding, spiking neural circuits that implement the complex gating mechanisms, and a store-and-release technique to enable neuron synchronization and faithful storage. While the presented techniques can be applied to map LSTM to any spiking neural network (SNN) simulator/emulator, here we choose the TrueNorth chip as the target platform by adhering to its hardware constraints. Three LSTMmore »
- Authors:
-
- Syracuse Univ., NY (United States)
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Publication Date:
- Research Org.:
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Sponsoring Org.:
- USDOE National Nuclear Security Administration (NNSA); National Science Foundation (NSF)
- OSTI Identifier:
- 1634300
- Report Number(s):
- LLNL-JRNL-810559
Journal ID: ISSN 2156-3357; 1017254
- Grant/Contract Number:
- AC52-07NA27344
- Resource Type:
- Accepted Manuscript
- Journal Name:
- IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Additional Journal Information:
- Journal Volume: 8; Journal Issue: 4; Journal ID: ISSN 2156-3357
- Publisher:
- IEEE
- Country of Publication:
- United States
- Language:
- English
- Subject:
- 97 MATHEMATICS AND COMPUTING; mathematics and computing; neural networks; neuromorphics; logic gates; computer architecture; recurrent neural networks; biological neural networks; synchronization; spiking neural networks; long short-term memory; neuromorphic hardware
Citation Formats
Shrestha, Amar, Ahmed, Khadeer, Wang, Yanzhi, Widemann, David P., Moody, Adam T., Van Essen, Brian C., and Qiu, Qinru. Modular Spiking Neural Circuits for Mapping Long Short-Term Memory on a Neurosynaptic Processor. United States: N. p., 2018.
Web. doi:10.1109/JETCAS.2018.2856117.
Shrestha, Amar, Ahmed, Khadeer, Wang, Yanzhi, Widemann, David P., Moody, Adam T., Van Essen, Brian C., & Qiu, Qinru. Modular Spiking Neural Circuits for Mapping Long Short-Term Memory on a Neurosynaptic Processor. United States. https://doi.org/10.1109/JETCAS.2018.2856117
Shrestha, Amar, Ahmed, Khadeer, Wang, Yanzhi, Widemann, David P., Moody, Adam T., Van Essen, Brian C., and Qiu, Qinru. Fri .
"Modular Spiking Neural Circuits for Mapping Long Short-Term Memory on a Neurosynaptic Processor". United States. https://doi.org/10.1109/JETCAS.2018.2856117. https://www.osti.gov/servlets/purl/1634300.
@article{osti_1634300,
title = {Modular Spiking Neural Circuits for Mapping Long Short-Term Memory on a Neurosynaptic Processor},
author = {Shrestha, Amar and Ahmed, Khadeer and Wang, Yanzhi and Widemann, David P. and Moody, Adam T. and Van Essen, Brian C. and Qiu, Qinru},
abstractNote = {Due to the distributed and asynchronous nature of neural computation through low-energy spikes, brain-inspired hardware systems offer high energy efficiency and massive parallelism. One such platform is the IBM TrueNorth neurosynaptic system. Recently, TrueNorth compatible representation learning algorithms have emerged, achieving close to the state-of-the-art performance in various data sets. However, its application in temporal sequence processing models, such as recurrent neural networks (RNNs), is still only at the proof of concept level. There is an inherent difficulty in capturing temporal dynamics of an RNN using spiking neurons, which is only exasperated by the hardware constraints in connectivity and synaptic weight resolution. This paper presents a design flow that overcomes these difficulties and maps a special case of recurrent networks called long short-term memory (LSTM) onto a spike-based platform. The framework utilizes various approximation techniques, such as activation discretization, weight quantization, and scaling and rounding, spiking neural circuits that implement the complex gating mechanisms, and a store-and-release technique to enable neuron synchronization and faithful storage. While the presented techniques can be applied to map LSTM to any spiking neural network (SNN) simulator/emulator, here we choose the TrueNorth chip as the target platform by adhering to its hardware constraints. Three LSTM applications, parity check, extended Reber grammar, and question classification, are evaluated. The tradeoffs among accuracy, performance, and energy tradeoffs achieved on TrueNorth are demonstrated. Finally, this is compared with the performance on an SNN platform without hardware constraints, which represents the upper bound of the achievable accuracy.},
doi = {10.1109/JETCAS.2018.2856117},
journal = {IEEE Journal on Emerging and Selected Topics in Circuits and Systems},
number = 4,
volume = 8,
place = {United States},
year = {Fri Jul 13 00:00:00 EDT 2018},
month = {Fri Jul 13 00:00:00 EDT 2018}
}
Web of Science
Figures / Tables:
Figures / Tables found in this record: