Spike Module

Spike

Spike mechanism implementation.

class lava.lib.dl.slayer.spike.spike.Spike(*args, **kwargs)

Bases: Function

Spiking mechanism with autograd link.

f_s(x) &= \mathcal{H}(x - \vartheta)

Parameters:

  • voltage (torch tensor) – neuron voltage.

  • threshold (float or torch tensor) – neuron threshold

  • tau_rho (float) – gradient relaxation

  • scale_rho (float) – gradient scale

  • graded_spike (bool) – flag for graded spike

  • voltage_last (torch tensor) – voltage at t=-1

  • scale (int) – variable scale value.

Returns:

spike tensor

Return type:

torch tensor

Examples

>>> spike = Spike.apply(v, th, tau_rho, scale_rho, False, 0, 1)
static backward(ctx, grad_spikes)
derivative = None
static forward(ctx, voltage, threshold, tau_rho, scale_rho, graded_spike, voltage_last, scale)

Complex Spike (Phase Threshold)

Complex spike (Phase threshold) implementation.

class lava.lib.dl.slayer.spike.complex.Spike(*args, **kwargs)

Bases: Function

Complex spike function with autograd link.

f_s(z) &= \mathcal{H}(|z| - \vartheta)\,\delta(\arg(z))

Parameters:

  • real (torch tensor) – real component of neuron response.

  • imag (torch tensor) – imaginary component of neuron response.

  • threshold (torch tensor or float) – neuron threshold.

  • tau_rho (float) – gradient relaxation

  • scale_rho (float) – gradient scale

  • graded_spike (bool) – flag for graded spike

  • imag_last (torch tensor) – imaginary response at t=-1

  • scale (int) – variable scale value.

Returns:

spike tensor

Return type:

torch tensor

Examples

>>> spike = Spike.apply(re, im, th, tau_rho, scale_rho, False, 0, 1)
static backward(ctx, grad_output)
derivative = None
static forward(ctx, real, imag, threshold, tau_rho, scale_rho, graded_spike, imag_last, scale)

Module contents

class lava.lib.dl.slayer.spike.Spike(*args, **kwargs)

Bases: Function

Spiking mechanism with autograd link.

f_s(x) &= \mathcal{H}(x - \vartheta)

Parameters:

  • voltage (torch tensor) – neuron voltage.

  • threshold (float or torch tensor) – neuron threshold

  • tau_rho (float) – gradient relaxation

  • scale_rho (float) – gradient scale

  • graded_spike (bool) – flag for graded spike

  • voltage_last (torch tensor) – voltage at t=-1

  • scale (int) – variable scale value.

Returns:

spike tensor

Return type:

torch tensor

Examples

>>> spike = Spike.apply(v, th, tau_rho, scale_rho, False, 0, 1)
static backward(ctx, grad_spikes)
derivative = None
static forward(ctx, voltage, threshold, tau_rho, scale_rho, graded_spike, voltage_last, scale)