COSYNE

Log in

With multi-electrode recording it is now possible to simultaneously record spikes from hundreds of neurons. Inferring the functional connectivity within these ensembles of neurons is an important question in neuroscience. To solve such tasks two pieces of information can be used. One is the likelihood function, that is, the information the recorded spikes carry about the connectivity. The other is the prior; from previous recordings we know that the network connectivity seems to be sparse and that interactions tend to unfold smoothly over time. Much recent progress has been made using a good likelihood function. Here we rather focus on the prior and use a simple linearized likelihood function in a maximum a posteriori framework. Using simulated data we show that our method can well infer connectivity patterns from spikes. In many cases it is at least as accurate as an approach that uses a very good likelihood function but no prior. We then proceed to test the method on spike data recorded from the motor cortex (area M1) of an awake monkey. The proposed method is conceptually simple and is efficient for large numbers of neurons and spikes. This method may thus be a useful tool for the analysis of data obtained from multi-unit recording techniques.