Building block of a programmable neuromorphic substrate: A digital neurosynaptic core

Arthur, John V.; Merolla, Paul; Akopyan, Filipp; Alvarez, Rodrigo; Cassidy, Andrew; Chandra, Shyamal; Esser, Steven K.; Imam, Nabil; Risk, W. P.; Rubin, Daniel B. D.; Manohar, Rajit; Modha, Dharmendra S.

doi:10.1109/ijcnn.2012.6252637

Cited by 111 publications

(67 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This approach is analogous to the manner in which software is used to program and compile computational processes in general purpose digital computers, except that the underlying neuromorphic hardware is radically different from digital ones in both system concept and electronic implementation. The approach is sufficiently general to be used on a wide range of electronic neural networks that have reconfigurable synaptic weights and reprogrammable connectivity (6,7,61).…”

Section: Discussionmentioning

confidence: 99%

Synthesizing cognition in neuromorphic electronic systems

Neftci

Binas

Rutishauser

et al. 2013

Proc. Natl. Acad. Sci. U.S.A.

130

122

View full text Add to dashboard Cite

The quest to implement intelligent processing in electronic neuromorphic systems lacks methods for achieving reliable behavioral dynamics on substrates of inherently imprecise and noisy neurons. Here we report a solution to this problem that involves first mapping an unreliable hardware layer of spiking silicon neurons into an abstract computational layer composed of generic reliable subnetworks of model neurons and then composing the target behavioral dynamics as a "soft state machine" running on these reliable subnets. In the first step, the neural networks of the abstract layer are realized on the hardware substrate by mapping the neuron circuit bias voltages to the model parameters. This mapping is obtained by an automatic method in which the electronic circuit biases are calibrated against the model parameters by a series of population activity measurements. The abstract computational layer is formed by configuring neural networks as generic soft winner-take-all subnetworks that provide reliable processing by virtue of their active gain, signal restoration, and multistability. The necessary states and transitions of the desired high-level behavior are then easily embedded in the computational layer by introducing only sparse connections between some neurons of the various subnets. We demonstrate this synthesis method for a neuromorphic sensory agent that performs real-time context-dependent classification of motion patterns observed by a silicon retina.decision making | sensorimotor | working memory | analog very large-scale integration | artificial neural systems U nlike digital simulations, in which the dynamics of neuronal models are encoded and calculated on general purpose digital hardware, "neuromorphic" emulations express the dynamics of the neural systems directly on an analogous physical substrate (1). Digital simulations have the advantage that they can be exactly and reliably programmed using numerical operations of very high precision. However, they suffer the disadvantage that they are cast on abstract binary electronic circuits whose operation is entirely divorced from the physical processes being simulated. Consequently, such simulations do not readily advance our understanding of how biological neural systems are able to attain their extraordinary physical performance, using only large numbers of apparently unreliable, slow, and imprecise neural components, a problem first recognized by von Neumann more than a half century ago (2). Furthermore, the reliability of digital systems comes at high cost of the circuit complexity necessary for the orchestration of communication and processing, an overhead that declares itself also in the costs of system construction and power dissipation (3).The alternative, neuromorphic, approach to information processing strives to capture in complementary metal-oxide semiconductor (CMOS) very large-scale integration (VLSI) electronic technology the more distributed, asynchronous, and limited precision nature of biological intelligent systems (1, 4).†, ‡ Research...

show abstract

Section: Discussionmentioning

confidence: 99%

Synthesizing cognition in neuromorphic electronic systems

Neftci

Binas

Rutishauser

et al. 2013

Proc. Natl. Acad. Sci. U.S.A.

130

122

View full text Add to dashboard Cite

show abstract

“…To this end, we have pursued neuroscience [42], [43], [9], nanotechnology [3], [4], [44], [5], [6], and supercomputing [11], [12]. Building on these insights, we are marching ahead to develop and demonstrate a novel, ultra-low power, compact, modular, non-von Neumann, cognitive computing architecture, namely, TrueNorth.…”

Section: Discussionmentioning

confidence: 99%

“…To this end, under the auspices of DARPA SyNAPSE, we are developing a novel, ultra-low power, compact, modular architecture called TrueNorth that consists of an interconnected and communicating network of extremely large numbers of neurosynaptic cores [3], [4], [5], [6]. Each core integrates computation (neurons), memory (synapses), and intra-core communication (axons), breaking the von Neumann bottleneck [7].…”

mentioning

confidence: 99%

Compass: A scalable simulator for an architecture for cognitive computing

Preissl

Wong

Datta

et al. 2012

2012 International Conference for High Performance Computing, Networking, Storage and Analysis

Self Cite

View full text Add to dashboard Cite

Abstract-Inspired by the function, power, and volume of the organic brain, we are developing TrueNorth, a novel modular, non-von Neumann, ultra-low power, compact architecture. TrueNorth consists of a scalable network of neurosynaptic cores, with each core containing neurons, dendrites, synapses, and axons. To set sail for TrueNorth, we developed Compass, a multi-threaded, massively parallel functional simulator and a parallel compiler that maps a network of long-distance pathways in the macaque monkey brain to TrueNorth. We demonstrate near-perfect weak scaling on a 16 rack IBM® Blue Gene®/Q (262144 CPUs, 256 TB memory), achieving an unprecedented scale of 256 million neurosynaptic cores containing 65 billion neurons and 16 trillion synapses running only 388× slower than real time with an average spiking rate of 8.1 Hz. By using emerging PGAS communication primitives, we also demonstrate 2× better real-time performance over MPI primitives on a 4 rack Blue Gene/P (16384 CPUs, 16 TB memory).I. INTRODUCTION The brain and modern computers have radically different architectures [1] suited for complementary applications. Modern computing posits a stored program model, traditionally implemented in digital, synchronous, serial, centralized, fast, hardwired, general-purpose, brittle circuits, with explicit memory addressing imposing a dichotomy between computation and data. In stark contrast, the brain uses replicated computational units of neurons and synapses implemented in mixedmode analog-digital, asynchronous, parallel, distributed, slow, reconfigurable, specialized, and fault-tolerant biological substrates, with implicit memory addressing blurring the boundary between computation and data [2]. It is therefore no surprise that one cannot emulate the function, power, volume, and realtime performance of the brain within the modern computer architecture. This task requires a radically novel architecture.Today, one must still build novel architectures in CMOS technology, which has evolved over the past half-century to serve modern computers and which is not optimized for delivering brain-like functionality in a compact, ultra-lowpower package. For example, biophysical richness of neurons and 3D physical wiring are out of the question at the very outset. We need to shift attention from neuroscientific richness that is sufficient to mathematical primitives that are necessary. A question of profound relevance to science, technology, business, government, and society is how closely can one approximate the function, power, volume, and real-time performance of the brain within the limits of modern technology.

show abstract

“…deSNN is also implementable on other recently proposed SNN chips of the same class, such as the digital IBM SNN chip (Merolla et al 2012) as well as on FPGA systems (Mitra et al, 2009). Despite the fast, on-pass learning in the deSNN models, in terms of large scale modelling of millions and billions of neurons using the SpiNNaker SNN supercomputer system (Jin et al, 2010) for simulation purposes would be appropriate, especially at the level of parameter optimisation.…”

Section: Neuromorphic Implementation Of Desnnmentioning

confidence: 99%

Dynamic evolving spiking neural networks for on-line spatio- and spectro-temporal pattern recognition

et al. 2013

View full text Add to dashboard Cite

On-line learning and recognition of spatio-and spectro-temporal data (SSTD) is a very challenging task and an important one for the future development of autonomous machine learning systems with broad applications. Models based on spiking neural networks (SNN) have already proved their potential in capturing spatial and temporal data. One class of them, the evolving SNN (eSNN), uses a one-pass rank-order learning mechanism and a strategy to evolve a new spiking neuron and new connections to learn new patterns from incoming data. So far these networks have been mainly used for fast image and speech frame-based recognition. Alternative spike-time learning methods, such as SpikeTiming Dependent Plasticity (STDP) and its variant Spike Driven Synaptic Plasticity (SDSP), can also be used to learn spatio-temporal representations, but they usually require many iterations in an unsupervised or semi-supervised mode of learning. This paper introduces a new class of eSNN, dynamic eSNN, that utilise both rank-order learning and dynamic synapses to learn SSTD in a fast, on-line mode. The paper also introduces a new model called deSNN, that utilises rank-order learning and SDSP spike-time learning in unsupervised, supervised, or semi-supervised modes. The SDSP learning is used to evolve dynamically the network changing connection weights that capture spatio-temporal spike data clusters both during training and during recall. The new deSNN model is first illustrated on simple examples and then applied on two case study applications: (1) moving object recognition using address-event representation (AER) with data collected using a silicon retina device; (2) EEG SSTD recognition for brain-computer interfaces. The deSNN models resulted in a superior performance in terms of accuracy and speed when compared with other SNN models that use either rank-order or STDP learning. The reason is that the deSNN makes use of both the information contained in the order of the first input spikes (which information is explicitly present in input data streams and would be crucial to consider in some tasks) and of the information contained in the timing of the following spikes that is learned by the dynamic synapses as a whole spatio-temporal pattern. The paper is published in Neural Networks, Elsevier, e- AbstractOn-line learning and recognition of spatio-and spectro-temporal data (SSTD) is a very challenging task and an important one for the future development of autonomous machine learning systems with broad applications. Models based on spiking neural networks (SNN) have already proved their potential in capturing spatial and temporal data. One class of them, the evolving SNN (eSNN), uses a one-pass rank-order learning mechanism and a strategy to evolve a new spiking neuron and new connections to learn new patterns from incoming data. So far these networks have been mainly used for fast image and speech frame-based recognition. Alternative spike-time learning methods, such as Spike-Timing Dependent Plasticity (STDP) and its variant S...

show abstract

Building block of a programmable neuromorphic substrate: A digital neurosynaptic core

Cited by 111 publications

References 21 publications

Synthesizing cognition in neuromorphic electronic systems

Synthesizing cognition in neuromorphic electronic systems

Compass: A scalable simulator for an architecture for cognitive computing

Dynamic evolving spiking neural networks for on-line spatio- and spectro-temporal pattern recognition

Contact Info

Product

Resources

About