Identification of time-varying neural dynamics from spiking activities using Chebyshev polynomials

Neural plasticity, elicited by processes such as development and learning, is an important biological attribute which can be viewed as a time-varying property of the nervous system. In this paper, we investigated the novel use of Chebyshev polynomials to estimate the changes in model parameters efficiently for time-varying dynamical systems with binary inputs and outputs. A forward orthogonal least square (FOLS) algorithm selected the significant model terms. Extensive simulations showed that the proposed algorithm identified the system changes more accurately in comparison with adaptive filter. This approach can be applied to identify not only gradual but also abrupt temporal evolutions of neural dynamics underlying nervous system activity with high sensitivity and accuracy by observing input and output spike trains only.