Digital hardware implementation of a stochastic two-dimensional neuron model. Issue 4 (November 2016)

• Record Type: Journal Article
• Authors: Grassia, F.; Kohno, T.; Levi, T.
• Is Part Of: Journal of physiology -Paris. Volume 110:Issue 4(2016)Part A
• Page Start: 409
• Page End: 416
• Publication Date: 2016-11
• Subjects: Neuromorphic engineering -- Stochastic neuron -- Spiking neuron model -- Noise -- FPGA silicon neuron; Physiology -- Periodicals; 571.105
• DOI: 10.1016/j.jphysparis.2017.02.002
• Languages: English
• ISSNs: 0928-4257

Abstract

Highlights: This study explores the feasibility of stochastic neuron simulation in FPGA. The stochasticity is added by a current noise using an Ornstein–Uhlenbeck process. The neuron model's computations are performed in arithmetic pipelines. The proposed FPGA implementation makes the silicon neuron more biologically plausible. Abstract: This study explores the feasibility of stochastic neuron simulation in digital systems (FPGA), which realizes an implementation of a two-dimensional neuron model. The stochasticity is added by a source of current noise in the silicon neuron using an Ornstein–Uhlenbeck process. This approach uses digital computation to emulate individual neuron behavior using fixed point arithmetic operation. The neuron model's computations are performed in arithmetic pipelines. It was designed in VHDL language and simulated prior to mapping in the FPGA. The experimental results confirmed the validity of the developed stochastic FPGA implementation, which makes the implementation of the silicon neuron more biologically plausible for future hybrid experiments.