Simulation of Large Scale Neural Models With Event-Driven Connectivity Generation.

Nathalie Azevedo Carvalho,Sylvain Contassot-Vivier,Laure Buhry,Dominique Martinez

doi:10.3389/fninf.2020.522000

Abstract

Accurate simulations of brain structures is a major problem in neuroscience. Many works are dedicated to design better models or to develop more efficient simulation schemes. In this paper, we propose a hybrid simulation scheme that combines time-stepping second-order integration of Hodgkin-Huxley (HH) type neurons with event-driven updating of the synaptic currents. As the HH model is a continuous model, there is no explicit spike events. Thus, in order to preserve the accuracy of the integration method, a spike detection algorithm is developed that accurately determines spike times. This approach allows us to regenerate the outgoing connections at each event, thereby avoiding the storage of the connectivity. Consequently, memory consumption is significantly reduced while preserving execution time and accuracy of the simulations, especially the spike times of detailed point neuron models. The efficiency of the method, implemented in the SiReNe software1, is demonstrated by the simulation of a striatum model which consists of more than 106 neurons and 108 synapses (each neuron has a fan-out of 504 post-synaptic neurons), under normal and Parkinson's conditions.

Highlights

Major projects such as Blue Brain (Markram, 2006), the Human Brain Project (Einevoll et al, 2019), the BRAIN Initiative or Mindscope (Hawrylycz et al, 2016), aim at simulating a brain or brain structures
We develop a spike detection method for HH neurons, that accurately determines the spike timings so that the accuracy of the second-order Runge-Kutta methods (RK2) is preserved when connectivity is generated at spiking events
We present the different types of simulation schemes: the common time-stepping approach, the event-driven connectivity generation, as well as the spike detection method we developed

Summary

Introduction

Major projects such as Blue Brain (Markram, 2006), the Human Brain Project (Einevoll et al, 2019), the BRAIN Initiative or Mindscope (Hawrylycz et al, 2016), aim at simulating a brain or brain structures. We propose a hybrid approach that combines a time-stepping approach for the numerical part of the simulation with an event-driven updating of the synaptic currents for complex HodgkinHuxley (HH) type neurons. This approach is well-suited to studies of real-time neural mechanisms requiring high accuracy, such as pathological behaviors like the Parkinson’s disease. Our event-driven approach completely avoids the storage of the connectivity pattern by regenerating the connectivity on the fly, when needed, after spiking events.

Methods

Results

Conclusion