Neurons

class snngrow.base.neuron.BaseNode.BaseNode(v_threshold: float = 1.0, v_reset: float = 0.0, surrogate_function: Callable = Sigmoid(alpha=4.0, spiking=True), detach_reset: bool = False, parallel_optim: bool = False, T: int = 1, spike_out: bool = False)

基类：Module

参数:

  • v_threshold (float) – threshold voltage

  • v_reset (float) – reset voltage. If not None, the neuron’s voltage will be set to v_reset after firing a spike. If None, the neuron’s voltage will subtract v_threshold after firing a spike

  • surrogate_function (Callable) – the function for calculating surrogate gradients of the heaviside step function in backward

  • detach_reset (bool) – detach the computation graph of reset in backward

  • parallel_optim (bool) – parallel optimization

  • T (int) – time steps

  • spike_out (bool) – whether to output SpikeTensor

The base class of differentiable spiking neurons.

detach()

Detach all stateful variables.

Tip

We can use this function to implement TBPTT(Truncated Back Propagation Through Time).

extra_repr()

Set the extra representation of the module

To print customized extra information, you should re-implement this method in your own modules. Both single-line and multi-line strings are acceptable.

forward(x: Tensor)

Defines the computation performed at every call.

Should be overridden by all subclasses.

备注

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

static hard_reset(v: Tensor, spike: Tensor, v_reset: float, spike_out: bool)
memories()
返回:

an iterator over all stateful variables

返回类型:

Iterator

named_memories()
返回:

an iterator over all stateful variables and their names

返回类型:

Iterator

abstract neuronal_dynamics(x: Tensor)

The neuronal dynamics difference equation. The sub-class must implement this function.

neuronal_fire(x: Tensor)

The neuronal fire difference equation.

neuronal_reset(spike)

The neuronal reset difference equation.

parallel_optim_forward(x_seq: Tensor)
参数:

x (torch.Tensor with ``shape = [T * N, *] ``) – input tensor with ``shape = [T * N, *] ``

The parallel forward function, which is implemented by calling simple_forward(x_seq[t]) over T times

register_memory(name: str, value)
参数:

  • name (str) – variable’s name

  • value (any) – variable’s value

Register the variable to memory dict, which saves stateful variables (e.g., the membrane potential of a spiking neuron). The reset value of this variable will be value. self.name will be set to value after each calling of self.reset().

reset()

Reset all stateful variables to their default values.

set_reset_value(name: str, value)
simple_forward(x: Tensor)
参数:

x (torch.Tensor) – increment of voltage inputted

返回:

out spikes

返回类型:

torch.Tensor

Forward by the order of dynamics - fire - reset.

static soft_reset(v: Tensor, spike: Tensor, v_threshold: float, spike_out: bool)
training: bool
v_float_to_tensor(x: Tensor)
class snngrow.base.neuron.LIFNode.LIFNode(tau: float = 2.0, decay_input: bool = True, v_threshold: float = 1.0, v_reset: float = 0.0, surrogate_function: Callable = Sigmoid(alpha=4.0, spiking=True), detach_reset: bool = False, parallel_optim: bool = False, T: int = 1, spike_out: bool = False)

基类：BaseNode

参数:

  • tau (float) – membrane time constant

  • decay_input (bool) – the input will decay

  • v_threshold (float) – threshold voltage

  • v_reset (float) – reset voltage. If not None, the neuron’s voltage will be set to v_reset after firing a spike. If None, the neuron’s voltage will subtract v_threshold after firing a spike

  • surrogate_function (Callable) – the function for calculating surrogate gradients of the heaviside step function in backward

  • detach_reset (bool) – detach the computation graph of reset in backward

  • parallel_optim (bool) – parallel optimization

  • T (int) – time steps

  • spike_out (bool) – whether to output SpikeTensor

The Leaky Integrate-and-Fire(LIF) neuron

extra_repr()

Set the extra representation of the module

To print customized extra information, you should re-implement this method in your own modules. Both single-line and multi-line strings are acceptable.

neuronal_dynamics(x: Tensor)

The neuronal dynamics difference equation. The sub-class must implement this function.

static neuronal_dynamics_decay_input(x: Tensor, v: Tensor, v_reset: float, tau: float)
static neuronal_dynamics_decay_input_reset0(x: Tensor, v: Tensor, tau: float)
static neuronal_dynamics_no_decay_input(x: Tensor, v: Tensor, v_reset: float, tau: float)
static neuronal_dynamics_no_decay_input_reset0(x: Tensor, v: Tensor, tau: float)
training: bool
class snngrow.base.neuron.IFNode.IFNode(v_threshold: float = 1.0, v_reset: float = 0.0, surrogate_function: Callable = Sigmoid(alpha=4.0, spiking=True), detach_reset: bool = False, parallel_optim: bool = False, T: int = 1, spike_out: bool = False)

基类：BaseNode

参数:

  • v_threshold (float) – threshold voltage

  • v_reset (float) – reset voltage. If not None, the neuron’s voltage will be set to v_reset after firing a spike. If None, the neuron’s voltage will subtract v_threshold after firing a spike

  • surrogate_function (Callable) – the function for calculating surrogate gradients of the heaviside step function in backward

  • detach_reset (bool) – detach the computation graph of reset in backward

  • parallel_optim (bool) – parallel optimization

  • T (int) – time steps

  • spike_out (bool) – whether to output SpikeTensor

The Integrate-and-Fire(IF) neuron, without decay input as LIF neuron.

neuronal_dynamics(x: Tensor)

The neuronal dynamics difference equation. The sub-class must implement this function.

training: bool