Skip to content

jaxfne TFNE-Izhikevich Single Emitter Explorer

Interactive browser-side preview for exploring reduced-emitter waveform dynamics before configuring a jaxfne scaffold.

Run status

artifact_class: interactive_tool
run_status: browser_euler_preview
model_status: parameter_intuition_only
field_solver_status: not_applicable
amplitude_status: native_unscaled
solver: browser_euler_dt_0.25ms

This tool runs entirely in the browser. It is a parameter intuition aid — not a jaxfne execution. All outputs are browser-computed previews for qualitative exploration only.

What this tool does

The explorer lets you sweep Izhikevich reduced-emitter parameters in real time and observe the resulting waveform, phase portrait, and nullcline structure. Eleven biological presets are included: E-Regular, E-Bursting, E-Chattering, E-Wide, PV, SST, VIP, FS, LTS, RZ (Resonator), and TC (Thalamocortical).

Use it to:

  1. Develop intuition for how a, b, c, d shape spike patterns before a jaxfne run.
  2. Compare E vs PV vs SST vs VIP waveform classes side by side.
  3. Understand the phase portrait (v vs u nullclines) for each cell type.
  4. Choose a starting drive level I before setting drive_gain in AGSDR tuning.

Transition to jaxfne

After exploring parameters here, use the package-native path for any actual simulation:

import jaxfne as jtfne

# Configure with the presets that match your exploration
cfg = jtfne.suite2_v1_v4_config(seed=42, n_per_area=80, duration_ms=1000.0, dt_ms=0.1)
model = jtfne.construct(cfg)
bundle = jtfne.suite2_run_bundle(model, seed=42, duration_ms=1000.0, dt_ms=0.1)

The emitter presets in the explorer correspond to jtfne.suite2_celltype_presets(). The drive level I maps to the drive_gain tunable parameter in jtfne.agsdr().

Scope

  • Browser Euler solver at dt=0.25 ms. Not the jaxfne JAX kernel.
  • Outputs are waveform previews only. No source projection, no field readout.
  • amplitude_status: native_unscaled — arbitrary units throughout.
  • No biological mechanism is implied by the parameter sweep.

Interactive panel

Browser-only preview

The panel below runs a forward-Euler Izhikevich solver in JavaScript. It requires no server, no Python, and no JAX. Adjust sliders to explore — then use jaxfne for any run that produces evidence.

run_status: browser_euler_preview
model_status: parameter_intuition_only
amplitude_status: native_unscaled