006 — PING E→I Coupling Video

Abstract

External drive is one knob (nb003); internal E→I→E coupling strength is the other. With the stim-window overdrive pinned at 1× (flat input through the trial), this notebook sweeps the E→I coupling strength — walking the network from the async baseline (E and I effectively decoupled) through the emergence of gamma as the feedback loop closes. Input rate and $W_\text{in}$ are bumped relative to nb003 / nb005 so E has enough baseline drive to recruit I at all without a stim pulse. See The Notebooks for how this entry’s runner/artifact/figure triple is wired up.

Methods

A scan over the oscilloscope video subcommand with stim-overdrive held at 1× and E→I strength sweeping 0 → 2 over 300 frames; input rate 200 Hz, $W_\text{in}$ init (1.8, 0.36). All knobs are hardcoded literals in src/pinglab/notebooks/nb006.py per the runner contract.

Results

Population rates

Discussion

TODO: discussion paragraph — write what the results above mean for the project.

Next steps

The recruitment threshold near ei ≈ 1.6 here is the practical floor for any trained PING circuit — at lower ei (e.g. 0.5) the network silently degenerates into feedforward-E with rate_i = 0 Hz. A natural follow-up is an independent 2D sweep over (ei-strength, input-rate) rendered as a heatmap of rate_i, since the threshold here is conditional on the bumped input drive used in this entry. With that map in hand, the trained-PING entries can choose an ei / drive pair from inside a verified PING basin instead of guessing.

Appendix

Reproduction

uv run src/pinglab/notebooks/nb006.py