004 — PING Input Rate Video
Abstract
Companion to nb003. Where nb003 holds the input rate fixed and sweeps an in-window overdrive multiplier, this entry has no stim window — the Poisson input rate is flat over the whole simulation, and the sweep walks that single rate from low to high. The aim is to map PING onset as a function of constant input drive, with no transient stimulus to confound the picture.
Methods
A scan over the oscilloscope video subcommand: each frame is a fresh 600 ms simulation on MNIST digit 0, sample 0, with the Poisson input rate set per frame from the scan and held flat across the whole run (overdrive=1, no stim elevation). The sweep walks the rate from 25 Hz → 1000 Hz over 300 frames. All knobs are hardcoded literals in src/pinglab/notebooks/nb004.py per the runner contract.
| Parameter | Value |
|---|---|
| Setup | |
| Model | ping |
| Input | mnist d0 s0 |
| Seed | 42 |
| Architecture | |
| N_E / N_I | 512 / 128 |
| dt / T | 0.1 / 600 ms |
| Stimulus & scan | |
| Envelope | flat (no stim window, overdrive=1) |
| Input-rate scan | 25–1000 Hz |
| Frames / FPS | 300 / 30 |
| Provenance | |
| Tier | large |
| Elapsed | 8m 07s |
| Run ID | r014 |
| Git SHA | ? |
Results
Each frame is a fresh 600 ms sim on MNIST digit 0, sample 0. The Poisson input rate is held flat across the run at the value shown on the frame. Sweep walks the rate from 25 Hz → 1000 Hz over 300 frames.
Population rates
| Population | Mean rate (Hz) |
|---|---|
| E | 12.2 |
| I | 28.0 |
Rates above are from the canonical replay at input rate = 250 Hz.
Discussion
Filled in once the run produces a video.
Implications
Filled in once findings are written.
Next steps
Cross-reference the input-rate threshold against nb003’s overdrive threshold — together they parameterise the (base, in-window) drive plane and identify whether PING onset is dominated by total drive or by stim-window contrast.
Appendix
Reproduction
uv run src/pinglab/notebooks/nb004.py