TY - JOUR

T1 - Analysis of spike-driven processes through attributable components

AU - Rotstein, Horacio G.

AU - Tabak, Esteban G.

N1 - Funding Information:
Acknowledgments. This work was partially supported by the National Science Foundation grants DMS-1608077 (HGR) and DMS-1715753 (EGT), by NIH MH060605 and by the Office of Naval Research (EGT). HGR is thankful to the Courant Institute of Mathematical Sciences at NYU. The authors are grateful to Farzan Nadim, Dirk Bucher and Sam McKenzie for useful discussions.
Funding Information:
This work was partially supported by the National Science Foundation grants DMS-1608077 (HGR) and DMS-1715753 (EGT), by NIH MH060605 and by the Office of Naval Research (EGT). HGR is thankful to the Courant Institute of Mathematical Sciences at NYU. The authors are grateful to Farzan Nadim, Dirk Bucher and Sam McKenzie for useful discussions.

PY - 2019

Y1 - 2019

N2 - Postsynaptic neuron activity at both the sub and suprathreshold level is analyzed through the combination of: (1) the numerical simulation of a simple leaky integrate-and-fire model forced by both constant frequency and Poisson-distributed presynaptic spike-trains,(2) the transformation of the model's response into sequences describing non-summation effects in subthreshold and the probability of spiking within a time-window in suprathreshold dynamics, (3) for constant frequency input, the analysis of these sequences through an autoregressive linear model, and (4) for non-uniform input, their analysis through attributable components. It is found that the attributable component methodology can reproduce the dynamics on testing data, effectively replacing the original dynamical model, and that the optimal order of both the autoregressive and the attributable component model, is an indicator of the relative strength of the underlying depression and facilitation mechanisms.

AB - Postsynaptic neuron activity at both the sub and suprathreshold level is analyzed through the combination of: (1) the numerical simulation of a simple leaky integrate-and-fire model forced by both constant frequency and Poisson-distributed presynaptic spike-trains,(2) the transformation of the model's response into sequences describing non-summation effects in subthreshold and the probability of spiking within a time-window in suprathreshold dynamics, (3) for constant frequency input, the analysis of these sequences through an autoregressive linear model, and (4) for non-uniform input, their analysis through attributable components. It is found that the attributable component methodology can reproduce the dynamics on testing data, effectively replacing the original dynamical model, and that the optimal order of both the autoregressive and the attributable component model, is an indicator of the relative strength of the underlying depression and facilitation mechanisms.

KW - Attributable components

KW - Dimensional reduction

KW - History-dependent processes

KW - Synaptic short-term plasticity

KW - Time series

UR - http://www.scopus.com/inward/record.url?scp=85077453642&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85077453642&partnerID=8YFLogxK

U2 - 10.4310/CMS.2019.v17.n5.a1

DO - 10.4310/CMS.2019.v17.n5.a1

M3 - Article

AN - SCOPUS:85077453642

VL - 17

SP - 1177

EP - 1192

JO - Communications in Mathematical Sciences

JF - Communications in Mathematical Sciences

SN - 1539-6746

IS - 5

ER -