Nowotny, T. | Search bioRxiv

bioRxiv2018

Brian2GeNN: a system for accelerating alarge variety of spiking neural networkswith graphics hardware

\"Brian\" is a popular Python-based simulator for spiking neural networks, commonly used in computational neuroscience. GeNN is a C++-based meta-compiler for accelerating spiking neural network simulations using consumer or high performance grade graphics processing units (GPUs). Here we introduce a new software package, Brian2GeNN, that connects the two systems so that users can make use of GeNN GPU acceleration when developing their models in Brian, without requiring any technical knowledge about GPUs, C++ or GeNN. The new Brian2GeNN software uses a pipeline of code generation to translate Brian scripts into C++ code that can be used as input to GeNN, and subsequently can be run on suitable NVIDIA GPU accelerators. From the users perspective, the entire pipeline is invoked by adding two simple lines to their Brian scripts. We have shown that using Brian2GeNN, typical models can run tens to hundreds of times faster than on CPU.

neuroscience↗

bioRxiv2017

Mixtures are more salient stimuli in olfaction

In their natural environment, animals often encounter complex mixtures of odours. It is an open question whether and how responses to complex mixtures of multi-component odours differ from those to simpler mixtures or single components. To approach this question, we built a full-size model of the early olfactory system of honeybees, which predicts responses to both single odorants and mixtures. The model is designed so that olfactory response patterns conform to the statistics derived from experimental data for a variety of their properties. It also takes into account several biophysical processes at a minimal level, including processes of chemical binding and activation in receptors, and spike generation and transmission in the antennal lobe network. We verify that key findings from other experimental data, not used in building the model, are reproduced in it. In particular, we replicate the strong correlation among receptor neurons and the weaker correlation among projection neurons observed in experimental data and show that this decorrelation is predominantly due to inhibition by interneurons. By simulation and mathematical analysis of our model, we demonstrate that the chemical processes of receptor binding and activation already lead to significant differences between the responses to mixtures and those to single component stimuli. On average, the response latency of olfactory receptor neurons at low stimulus concentrations is reduced and the response patterns become less variable across concentrations as the number of odour components in the stimulus increases. These effects are preserved in the projection neurons. Our results suggest that the early olfactory system of insects may be particularly efficient in processing mixtures, which corresponds well to the observation that chemical signalling in nature predominantly involves mixtures.

neuroscience↗