Synaptic Classification via Spike-Triggered Extrapolation

TL;DR

This paper presents a novel statistical method for inferring neuronal synaptic connections from spike train data, effectively classifying excitatory, inhibitory, or null links even in noisy, complex networks.

Contribution

It introduces a Macro-Micro Extrapolation algorithm with a Spike-Triggered Estimator that improves inference accuracy in sparse data conditions.

Findings

Classifier accurately identifies synapses without error across noise levels.

Method works effectively even with observation windows larger than theoretical bounds.

Framework successfully distinguishes excitatory, inhibitory, and null connections.

Abstract

This work introduces a statistical procedure to infer the interaction graph of neuronal networks modeled by Galves-L\"ocherbach dynamics. The methodology performs bivariate inference, identifying synaptic links from the spike trains of pairs of neurons without observing the rest of the network. We propose a Macro-Micro Extrapolation algorithm to address data sparsity by inferring interactions in the limit $\Delta \to 0^+$. The core component is a Spike-Triggered Estimator that leverages the local reset property to decouple synaptic jumps from background noise. By employing an adaptive logic that switches between sample averaging and Pyramid Extrapolation, the framework categorizes connections as excitatory, inhibitory, or null. Numerical simulations demonstrate that the classifier identifies synapses without error across varying noise regimes and complex network topologies, even for observation windows broader than those predicted by the current theoretical bounds.