bioRxiv · 2026 · Preprint

A Kernel-Spectral Index for Single-Neuron Input-Output Complexity

Zixuan Huang, Wenjie Chen, Bo Lei, Guozhang Chen

The Spectral Complexity Index provides an interpretable, white-box kernel measure of single-neuron input-output computation at a fraction of surrogate-model cost.

The scientific question

How can single-neuron computational complexity be quantified without black-box surrogate networks?

Cortical pyramidal neurons perform deep nonlinear input-output transformations, but measuring their complexity by fitting multi-layer artificial neural networks is computationally costly and difficult to interpret.

Overview of the Spectral Complexity Index.
Figure 1 from the preprint ↗

The core idea

White-Box Kernel-Spectral Complexity Measure

We introduce the Spectral Complexity Index (SCI), which embeds presynaptic spike histories into an explicit kernel geometry and counts the ordered spectral modes needed to explain somatic responses.

What we found

Fast Cross-Species Separation and Stability

Across 35 biophysical human and mouse cortical neuron models, SCI separates the two species, aligns closely with deep-network complexity rankings, remains robust to perturbations, and cuts runtime tenfold.

Continue the question

Related reading

Explore our research

Cite this work

Zixuan Huang, Wenjie Chen, Bo Lei, Guozhang Chen. A Kernel-Spectral Index for Single-Neuron Input-Output Complexity. bioRxiv, 2026. doi:10.64898/2026.11.04.764153.