A Pulse-gated, Neural Implementation of the Backpropagation Algorithm

Sornborger, Andrew; Tao, Louis; Snyder, Jordan; Zlotnik, Anatoly

doi:10.1145/3320288.3320305

Cited by 8 publications

(6 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the basis of modern deep neural networks remains the error backpropagation algorithm [1], which though founded in rigorous mathematical optimization theory, has not been successfully demonstrated in a neurophysiologically realistic circuit. In a recent study, we proposed a neuromorphic architecture for learning that tunes the propagation of information forward and backwards through network layers using an endogenous timing mechanism controlled by thresholding of intensities [2]. This mechanism was demonstrated in simulation of analog currents, which represent the mean fields of spiking neuron populations.…”

Section: Discussionmentioning

confidence: 99%

“…In this study, we describe the hardware implementation of the backpropagation algorithm that makes use of mechanisms that have been developed and tested in simulation by the authors during the past decade, and synthesized in our recent study [2]. These proven neuronal and network features include propagation of graded information in a circuit composed of neural populations using synfiregated synfire chains (SGSCs) [11ś14], decision-making based on the interaction of synfire-chains [12], and regulation of Hebbian learning using pulse-gating [15,16].…”

Section: Contributionsmentioning

confidence: 99%

See 1 more Smart Citation

Implementing Backpropagation for Learning on Neuromorphic Spiking Hardware

Renner

Sheldon

Zlotnik

et al. 2020

Proceedings of the Neuro-Inspired Computational Elements Workshop

Self Cite

View full text Add to dashboard Cite

Section: Discussionmentioning

confidence: 99%

Section: Contributionsmentioning

confidence: 99%

Implementing Backpropagation for Learning on Neuromorphic Spiking Hardware

Renner

Sheldon

Zlotnik

et al. 2020

Proceedings of the Neuro-Inspired Computational Elements Workshop

Self Cite

View full text Add to dashboard Cite

“…In this study, we describe a hardware implementation of the backpropagation algorithm that addresses each of the issues (a)-(e) introduced above using a set of mechanisms that have been developed and tested in simulation by the authors during the past decade, synthesized in our recent study [55], and simplified and adapted here to the features and constraints of the Loihi chip. These neuronal and network features include propagation of graded information in a circuit composed of neural populations using synfire-gated synfire chains (SGSCs) [56][57][58][59], control flow based on the interaction of synfire-chains [57], and regulation of Hebbian learning using pulse-gating [60,61].…”

Section: Our Contributionmentioning

confidence: 99%

“…These neuronal and network features include propagation of graded information in a circuit composed of neural populations using synfire-gated synfire chains (SGSCs) [56][57][58][59], control flow based on the interaction of synfire-chains [57], and regulation of Hebbian learning using pulse-gating [60,61]. We simplify our previously proposed network architecture [55], and streamline its function. We demonstrate our approach using a proof-of-principle implementation on Loihi [5], and examine the performance of the algorithm for learning and inference of the MNIST test data set [62].…”

Section: Our Contributionmentioning

confidence: 99%

The Backpropagation Algorithm Implemented on Spiking Neuromorphic Hardware

Renner¹,

Sheldon²,

Zlotnik³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

The capabilities of natural neural systems have inspired new generations of machine learning algorithms as well as neuromorphic very large-scale integrated (VLSI) circuits capable of fast, low-power information processing. However, most modern machine learning algorithms are not neurophysiologically plausible and thus are not directly implementable in neuromorphic hardware. In particular, the workhorse of modern deep learning, the backpropagation algorithm, has proven difficult to translate to neuromorphic hardware. In this study, we present a neuromorphic, spiking backpropagation algorithm based on pulse-gated dynamical information coordination and processing, implemented on Intel's Loihi neuromorphic research processor. We demonstrate a proof-of-principle three-layer circuit that learns to classify digits from the MNIST dataset. This implementation shows a path for using massively parallel, low-power, low-latency neuromorphic processors in modern deep learning applications.

show abstract

“…The first class of learning rules are gradient-based methods. They approximate backpropagation with various levels of biological plausibility [3,23,25,26,28,30,37,39,44]. From this category, we study the e-prop algorithm [6] in detail and provide a complete reimplementation.…”

Section: Introductionmentioning

confidence: 99%

Neko: a Library for Exploring Neuromorphic Learning Rules

Zhao

Wycoff

Getty

et al. 2021

Preprint

View full text Add to dashboard Cite

The field of neuromorphic computing is in a period of active exploration. While many tools have been developed to simulate neuronal dynamics or convert deep networks to spiking models, general software libraries for learning rules remain underexplored. This is partly due to the diverse, challenging nature of efforts to design new learning rules, which range from encoding methods to gradient approximations, from population approaches that mimic the Bayesian brain to constrained learning algorithms deployed on memristor crossbars. To address this gap, we present Neko, a modular, extensible library with a focus on aiding the design of new learning algorithms. We demonstrate the utility of Neko in three exemplar cases: online local learning, probabilistic learning, and analog on-device learning. Our results show that Neko can replicate the state-of-the-art algorithms and, in one case, lead to significant outperformance in accuracy and speed. Further, it offers tools including gradient comparison that can help develop new algorithmic variants. Neko is an open source Python library that supports PyTorch and TensorFlow backends.

show abstract

A Pulse-gated, Neural Implementation of the Backpropagation Algorithm

Cited by 8 publications

References 22 publications

Implementing Backpropagation for Learning on Neuromorphic Spiking Hardware

Implementing Backpropagation for Learning on Neuromorphic Spiking Hardware

The Backpropagation Algorithm Implemented on Spiking Neuromorphic Hardware

Neko: a Library for Exploring Neuromorphic Learning Rules

Contact Info

Product

Resources

About