Library for fitting a population of doubly-stochastic integrate-and-fire neurons to spike train data. The model accounts for fast independent and slower shared input fluctuations that dominate the low-dimensional collective dynamics. In particular, each neuron is driven by an independent Gaussian white noise process, whose mean varies according to a slower stochastic process that is shared among the population. The statistical inference method is described in: Donner, Opper, Ladenbauer, Inferring the collective dynamics of neuronal populations from single-trial spike trains using mechanistic models (under review)
An example is given in the file example.py, which includes generation of synthetic data from the generative model, parameter estimation, and visualization of the results. The code was written with Python 2.7.
Unreasonably small neuronal interspike intervals (ISIs < 3ms), due to spike sorting errors from in-vivo recordings for example, may cause problems. In this case we recommend to remove very small ISIs and/or set the parameter sorting_error to a larger value.
Required dependencies are: numpy, numba, scipy, multiprocessing, functools, warnings.
The code was developed by Christian Donner and Josef Ladenbauer. For technical questions please contact christian.research(at)mailbox.org.