Oscillators
A flat activation score gives one number per frame: how active it is right now. That’s enough to rank what a query returns, but it can’t express the thing this work is about. Two frames can both be active without being active together. A single number can’t tell those apart.
So we model each frame as an oscillator — something that pulses, carrying a phase (a position in a cycle) on top of its amplitude. Phase means nothing for one frame alone; it only means something relative to other frames. That relative reading is the point. Two frames sharing a phase are co-active in the strong sense, not just both switched on. Convergence, lock-in, drift — the regimes substrate dynamics tries to read — are all synchronization phenomena, things falling into or out of step. Modelling frames as oscillators chooses a representation where synchronization is the native quantity, and borrows the body of oscillator mathematics (and the brain models built on it) that already knows how to measure it.
What it gives the substrate
- Coherence. Because frames carry phase, we can ask how in-step a region is right now — one reading of how aligned a set of frames has become. The aim is to read this over a membrane, not over the whole substrate, where it averages out.
- Self-forming structure. Phase plus coupling lets clusters assemble themselves: frames used together are pulled toward the same rhythm. Nothing clusters the graph by hand; pulsing-together does it.
- A grip for the dynamics. Synchronization is a measured quantity in the oscillator literature, which is where the candidate readings for convergence, lock-in, and drift come from.