NeuroFlow: A General Purpose Spiking Neural Network Simulation Platform using Customizable Processors

Cheung, Kit; Schultz, Simon R.; Luk, Wayne

doi:10.3389/fnins.2015.00516

Cited by 74 publications

(38 citation statements)

References 60 publications

(85 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The modular structure of the code generation framework is also designed to be proof against future trends in both high performance computing and computational neuroscience research. Increasingly, high performance scientific computing relies on the use of heterogeneous computing architectures such as GPUs, FPGAs, and even more specialised hardware (Fidjeland et al, 2009;Richert et al, 2011;Brette & Goodman, 2012;Moore et al, 2012;Furber et al, 2014;Cheung et al, 2016), as well as techniques such as approximate computing (Mittal, 2016). In addition to the existing standalone mode, it is possible to write plugins for Brian to generate code for these platforms and techniques without modifying the core code, and there are several ongoing projects to do so.…”

Section: Discussionmentioning

confidence: 99%

Brian 2: an intuitive and efficient neural simulator

Stimberg

Brette

Goodman

2019

Preprint

113

View full text Add to dashboard Cite

To be maximally useful for neuroscience research, neural simulators must make it possible to define original models. This is especially important because a computational experiment might not only need descriptions of neurons and synapses, but also models of interactions with the environment (e.g. muscles), or the environment itself. To preserve high performance when defining new models, current simulators offer two options: low-level programming, or mark-up languages (and other domain specific languages). The first option requires time and expertise, is prone to errors, and contributes to problems with reproducibility and replicability. The second option has limited scope, since it can only describe the range of neural models covered by the ontology. Other aspects of a computational experiment, such as the stimulation protocol, cannot be expressed within this framework. "Brian" 2 is a complete rewrite of Brian that addresses this issue by using runtime code generation with a procedural equation-oriented approach. Brian 2 enables scientists to write code that is particularly simple and concise, closely matching the way they conceptualise their models, while the technique of runtime code generation automatically transforms high level descriptions of models into efficient low level code tailored to different hardware (e.g. CPU or GPU). We illustrate it with several challenging examples: a plastic model of the pyloric network of crustaceans, a closed-loop sensorimotor model, programmatic exploration of a neuron model, and an auditory model with real-time input from a microphone. from brian2 import * 2 defaultclock.dt = 0.01*ms; 3 Delta_T = 17.5*mV ; v_T = -40*mV ; tau = 2*ms ; tau_adapt = .02*second 4 tau_Ca = 150*ms ; tau_x = 2*second ; v_r = -68*mV ; tau_z = 5*second 5

show abstract

Section: Discussionmentioning

confidence: 99%

Brian 2: an intuitive and efficient neural simulator

Stimberg

Brette

Goodman

2019

Preprint

113

View full text Add to dashboard Cite

show abstract

“…Carlson et al [34] presented a simulation environment for large-scale spiking neural nets with evolutionary parameter tuning which harnessed the processing power of GPUs. More recently, Cheung et al [35] developed "NeuroFlow", a scalable platform for spiking neural nets on FPGA. Their system could simulate upto 400,000 neurons in real-time with a speedup of 2.83 times than that of GPUs.…”

Section: Gpus and Dspsmentioning

confidence: 99%

Recent trends in neuromorphic engineering

Soman

Jayadeva

2016

Big Data Anal

View full text Add to dashboard Cite

Neuromorphic Engineering has emerged as an exciting research area, primarily owing to the paradigm shift from conventional computing architectures to data-driven, cognitive computing. There is a diversity of work in the literature pertaining to neuromorphic systems, devices and circuits. This review looks at recent trends in neuromorphic engineering and its sub-domains, with an attempt to identify key research directions that would assume significance in the future. We hope that this review would serve as a handy reference to both beginners and experts, and provide a glimpse into the broad spectrum of applications of neuromorphic hardware and algorithms. Our survey indicates that neuromorphic engineering holds a promising future, particularly with growing data volumes, and the imminent need for intelligent, versatile computing.

show abstract

“…However, Neb can be extended to cover a range of other interesting areas. For example, in simulating the spiking of neurons, a task tackled by tools such as NeuroFlow [5]. Fig.…”

Section: Further Considerationsmentioning

confidence: 99%

A Domain Specific Language for accelerated Multilevel Monte Carlo simulations

Lindsey

Leslie

Luk

2016

2016 IEEE 27th International Conference on Application-Specific Systems, Architectures and Processors (ASAP)

Self Cite

View full text Add to dashboard Cite

Abstract-Monte Carlo simulations are used to tackle a wide range of exciting and complex problems, such as option pricing and biophotonic modelling. Since Monte Carlo simulations are both computationally expensive and highly parallelizable, they are ideally suited for acceleration through GPUs and FPGAs. Alongside these accelerators, Multilevel Monte Carlo techniques can be harnessed to further hasten simulations. However, researchers and application developers must invest a great deal of effort to design, optimise and test such Monte Carlo simulations. Furthermore, these models often have to be rewritten from scratch to target new hardware accelerators. This paper presents Neb, a Domain Specific Language for describing and generating Multilevel Monte Carlo simulations for a variety of hardware architectures. Neb compiles equations written in L A T E X to C++, OpenCL or Maxeler's MaxJ language, allowing acceleration through GPUs or FPGAs. Neb can be used to solve stochastic equations or to generate paths for analysis with other tools. To evaluate the performance of Neb, a variety of financial models are executed on CPUs, GPUs and FPGAs, demonstrating peak acceleration of 3.7 times with FPGAs in 40nm transistor technology, and 14.4 times with GPUs in 28nm transistor technology. Furthermore, the energy efficiency of these accelerators is compared, revealing FPGAs to be 8.73 times and GPUs 2.52 times more efficient than CPUs.

show abstract

NeuroFlow: A General Purpose Spiking Neural Network Simulation Platform using Customizable Processors

Cited by 74 publications

References 60 publications

Brian 2: an intuitive and efficient neural simulator

Brian 2: an intuitive and efficient neural simulator

Recent trends in neuromorphic engineering

A Domain Specific Language for accelerated Multilevel Monte Carlo simulations

Contact Info

Product

Resources

About