TemporalEncoder

Contents

TemporalEncoder#

class braintools.spike.TemporalEncoder(n_patterns, pattern_length=10, jitter=0.1)#

Encode temporal patterns using synchronized spike timing.

Encodes a sequence of integer symbols. Each symbol occupies a slot of pattern_length time steps; during that slot the neuron with the symbol’s index fires a fixed pattern of spikes (about 30 % of the slot, shifted by twice the symbol index), with Gaussian jitter. Symbols outside [0, n_patterns) leave their slot silent.

Parameters:

  • n_patterns (int) – Number of distinct symbols (and output neurons).

  • pattern_length (int) – Length of each temporal pattern in time steps.

  • jitter (float) – Standard deviation of the temporal jitter added to spike times, as a fraction of pattern_length.

Examples

>>> import brainstate
>>> import jax.numpy as jnp
>>> import braintools
>>> brainstate.random.seed(0)
>>> encoder = braintools.spike.TemporalEncoder(n_patterns=3, pattern_length=10, jitter=0.0)
>>> sequence = jnp.array([0, 1, 2, 1, 0])
>>> spikes = encoder(sequence)
>>> spikes.shape
(50, 3)
>>> # The active neuron in each slot decodes the symbol.
>>> jnp.argmax(spikes.reshape(5, 10, 3).sum(axis=1), axis=1)
Array([0, 1, 2, 1, 0], dtype=int32)
Contents