SpikeStream.jl

Spike-stream feature extraction for spiking neural systems

---

SpikeStream.jl is focused on feature extraction from spike-event streams.

Core Features

• spike_count(spike_times; t_start, t_end)
• spike_density(spike_times; t_start, t_end)
• isi_stats(spike_times)
• detect_bursts(spike_times; max_isi, min_spikes)
• windowed_spike_features(spike_times; window_size, step)
• normalized_feature_vector(spike_times)

Output Ranges

• spike_count → integer >= 0
• spike_density → real >= 0
• isi_stats → all fields non-negative
• detect_bursts → vector of index ranges (possibly empty)
• windowed_spike_features:
  • count >= 0
  • density >= 0
  • isi_mean >= 0
  • isi_cv >= 0
  • burst_count >= 0
• normalized_feature_vector → length-4 vector in [0, 1]

Package Boundary

SpikeStream.jl owns spike-stream feature extraction only. Kinetic / signal-dynamics APIs live in the Rust sibling Limen-Neural/kinetic-signals.

Area	Owner	Notes
spike_count, spike_density, isi_stats, detect_bursts, windowed_spike_features, normalized_feature_vector	SpikeStream.jl	Public API of this package
Hurst exponent	kinetic-signals	Formerly transitional compute_hurst (removed)
Hawkes intensity	kinetic-signals	Formerly transitional compute_hawkes (removed)
Surprise / geometric Brownian motion (GBM)	kinetic-signals	Formerly transitional compute_gbm_surprise (removed)
Entropy	kinetic-signals	Not part of SpikeStream.jl
Volatility	kinetic-signals	Not part of SpikeStream.jl

Integration

• No foreign-function interface (FFI) between SpikeStream.jl and kinetic-signals today.
• Spike fixtures for cross-package tests live in this package: test/fixtures/spike_vectors.json (LIM-41).
• Window note: with explicit t_end, windowed_spike_features uses half-open windows [t_start, t_end); spike_count / spike_density use inclusive ends. Fixtures encode this.
• kinetic-signals shared_vectors.json is Rust-only (not consumed by SpikeStream.jl).

Quick Start

using SpikeStream

spike_times = [0.001, 0.005, 0.009, 0.040, 0.042, 0.044, 0.090]

count = spike_count(spike_times)
density = spike_density(spike_times; t_start=0.0, t_end=0.1)
stats = isi_stats(spike_times)
bursts = detect_bursts(spike_times; max_isi=0.004, min_spikes=3)
windows = windowed_spike_features(spike_times; window_size=0.03, step=0.03)
vec = normalized_feature_vector(spike_times; t_start=0.0, t_end=0.1, max_density=200.0)

Installation

using Pkg
Pkg.add("SpikeStream")

License

Licensed under either of:

• MIT License (LICENSE-MIT)
• Apache License 2.0 (LICENSE-APACHE)

at your option.