API reference

The public modules of sparx, each documented from its source.

Module
sparx	Sparx: spiking neural networks in JAX and Flax, trained, simulated and served through dew.
sparx.nn	Flax linen layers for spiking networks, over time-major inputs [T, ...].
sparx.models	Spiking network architectures built from sparx.nn layers.
sparx.surrogate	The spike nonlinearity and the gradients it trains with.
sparx.encode	Turn a batch field into the time-major input [T, B, ...] of a spiking network.
sparx.losses	Differentiable losses over a network’s time-major outputs [T, B, ...].
sparx.rates	Read and regularize the firing rates spiking layers sow.
sparx.dynamics	Neuron, synapse and plasticity models, and the runner that scans them over time.
sparx.graph	Circuits and connectomes: populations of neurons joined by projections, simulated on one clock.
sparx.spiketrains	Statistics of and distances between spike trains, for comparing networks that cannot match spike for spike.
sparx.learn	Learning rules beyond surrogate-gradient backpropagation through time (design.md section 7).
sparx.objectives	Spiking networks as dew objectives, trained by dew’s Trainer.
sparx.metrics	Metrics over a spiking objective’s evaluation, which dew’s Trainer.fit(metrics=...) takes.
sparx.tasks	Trained spiking networks as inference tasks, which dew.pipeline loads from a run.
sparx.config	The run a spiking classifier trains as, one typed record that run.json holds.
sparx.datasets	Neuromorphic datasets as dense, binned spike counts.
sparx.serve	Serving streaming spiking models: many sessions, each with its own neuron state, in one batched program.
sparx.nir	Exchanging networks through NIR, the Neuromorphic Intermediate Representation (Pedersen et al. 2024).