braintools.spike module

Spike data: the SpikeTrains container, encoders and bitwise operations.

• SpikeTrains stores the spike times of a population in a padded, fixed-shape layout that works under jax.jit and jax.vmap, and converts to and from dense rasters, flat event lists and ragged lists.

• The encoders (LatencyEncoder, PoissonEncoder, …) turn numbers or time series into spike rasters.

• The bitwise functions (bitwise_or(), bitwise_and(), …) combine spike tensors element-wise and stay differentiable for surrogate-gradient spikes.

Spike data in one place: a fixed-shape container for spike times, encoders that turn numbers or time series into spike rasters, and element-wise bitwise operations on spike tensors.

Spike Trains

SpikeTrains stores the spike times of a population as a padded (n_neurons, capacity) array. It is a JAX pytree, so it can be built and used inside jax.jit and jax.vmap, and it converts to and from dense rasters (from_dense / to_dense), flat event lists (from_events / to_events) and ragged per-neuron lists (from_list / to_list). The braintools.visualize spike plots accept it directly.

SpikeTrains

Spike times of a population in a padded, fixed-shape layout.

Spike Encoders

Encoders that convert continuous-valued signals into spike trains (or, for StepCurrentEncoder, into input currents) for spiking network simulations. Before 0.4.0 they lived at the top level (braintools.RateEncoder and so on); those names are deprecated aliases that will be removed in 1.0.0.

LatencyEncoder	Encode input intensities as the time of a single spike (latency coding).
RateEncoder	Encode analog values into spike rates using various rate encoding methods.
PoissonEncoder	Encode inputs as Poisson spike trains.
PopulationEncoder	Encode scalar values using population coding.
BernoulliEncoder	Encode inputs using independent Bernoulli processes.
DeltaEncoder	Encode a time series by send-on-delta (level-crossing) modulation.
StepCurrentEncoder	Encode inputs as step current injections for LIF neurons.
SpikeCountEncoder	Encode inputs as exact spike counts over time windows.
TemporalEncoder	Encode temporal patterns using synchronized spike timing.
RankOrderEncoder	Encode inputs using rank order coding.

Bitwise Operations

Element-wise logical operations on spike tensors. Boolean inputs give boolean results; numeric inputs use arithmetic identities that keep the dtype and stay differentiable for surrogate-gradient spikes. The former top-level names (braintools.spike_bitwise_or and so on) remain available as aliases.

bitwise_or	Perform a bitwise OR operation on spike tensors.
bitwise_and	Perform a bitwise AND operation on spike tensors.
bitwise_iand	Perform a bitwise IAND (Inverse AND) operation on spike tensors.
bitwise_not	Perform a bitwise NOT operation on spike tensors.
bitwise_xor	Perform a bitwise XOR operation on spike tensors.
bitwise_ixor	Perform a bitwise IXOR (Inverse XOR) operation on spike tensors.
bitwise	Perform bitwise operations on spike tensors.