Accepted
On the teaching side a kernel is chosen to demonstrate how it shapes the output (see FOUNDATIONS.md §2, Tab 1). Options considered for where that kernel comes from:
- A library of canonical shapes.
- The same shapes with live parameters.
- Freehand drawn or typed arbitrary kernels.
Separately, the calcium indicator response — a fast rise plus exponential decay, parameterized by tau_rise / tau_decay — is a shape researchers actually care about. Offering it as a library entry bridges the abstract teaching tabs to the real Tab 2/3 calcium use case (FOUNDATIONS.md §1).
The teaching-side kernel is a parameterized library of canonical shapes — e.g. Gaussian, exponential decay, boxcar / moving-average, and the calcium indicator rise/decay shape (difference-of-Gaussians optional).
Each shape exposes live parameters (width, tau, length) via sliders, so changing them updates the output in real time. This is where the interactive teaching value lives.
Freehand drawn or typed kernels are explicitly excluded — not v1, not planned. They add flexibility a learner rarely needs and don't teach a named concept the way a labeled, parameterized shape does.
Pros
- Parameterized shapes are the interactive lesson: drag the width, watch the output smear.
- The calcium kernel as a library entry is the bridge from teaching to the flagship research use case.
- Excluding freehand keeps the UI and the mental model clean.
Cons
- No arbitrary kernel shapes on the teaching side. Acceptable, since every teaching goal is served by a named parameterized shape.
Note
- This ADR concerns kernels chosen as input (Tabs 1 and 3). Tab 2 does not take a kernel as input — it recovers one — so this ADR does not govern Tab 2's kernel.
See FOUNDATIONS.md §1 (calcium use case) and §2 (tab roles), and open question §10.4 (kernel source for the teaching side).