Recorders

Spike train recording and analysis utilities.

• BitstreamSpikeRecorder — records a 1D binary spike train (0/1 per timestep), returns a uint8 NumPy view, counts spikes, computes mean firing rate from dt_ms, and builds inter-spike-interval histograms.

from sc_neurocore import BitstreamSpikeRecorder

recorder = BitstreamSpikeRecorder(dt_ms=1.0)
for _ in range(1000):
    spike = neuron.step(current)
    recorder.record(spike)

print(f"Total spikes: {recorder.total_spikes()}")
print(f"Firing rate: {recorder.firing_rate_hz():.1f} Hz")

The recorder rejects non-binary spikes, negative dt_ms values, and non-positive histogram bin counts. dt_ms=0.0 is still accepted for legacy dry-run callers and returns a firing rate of 0.0.

Polyglot Mirrors

The Python API is the canonical recorder surface. Parity helper surfaces are kept in accel/julia/recorders/spike_recorder.jl, accel/rust/safety/spike_recorder.rs, and accel/mojo/kernels/spike_recorder.mojo for backend validation and low-level firing-rate/ISI helpers. The Rust safety mirror owns crate-level unit tests; the Julia and Mojo mirrors are syntax-checked in the recorder maintenance lane.

::: sc_neurocore.recorders.spike_recorder.BitstreamSpikeRecorder options: show_root_heading: true