We develop a simulator for spiking neural networks named Brian, with Dan Goodman (Goodman & Brette 2008; 2009; Brette & Goodman, 2009; Goodman & Brette 2013a; 2013b). The motto of Brian is “a simulator should not only save the time of processors, but also the time of scientists” (Brette 2012). It is written in Python and equation-oriented (Stimberg et al. 2014): users define models themselves by providing their mathematical definition (as opposed to built-in models). This choice makes Brian ideally suited for rapid model writing and for teaching. The first version of Brian relied on vectorised algorithms to make simulations efficient for large networks (Brette & Goodman 2011). The new version of Brian (2.0) relies on code generation to simulate models very efficiently on various platforms.

We have also developed a toolbox for Brian to fit spiking models to electrophysiological recordings (Rossant et al 2010; 2011), as well as an auditory toolbox (Fontaine et al 2011).

Our current projects include: running Brian on diverse platforms such as GPU (Brette & Goodman, 2012), FPGA and clusters, using Brian to run electrophysiology experiments, and further develop multicompartmental modeling in Brian (which is already in Brian 2.0).

Further information