hxtorch.spiking.transforms.encode.CoordinatesToSpikes

class hxtorch.spiking.transforms.encode.CoordinatesToSpikes(seq_length: int, t_early: float, t_late: float, dt: float = 1e-06, t_bias: Optional[float] = None)

Bases: torch.nn.modules.module.Module

Convert values between 0 and 1 to spikes in a given timeframe

__init__(seq_length: int, t_early: float, t_late: float, dt: float = 1e-06, t_bias: Optional[float] = None) → None

Construct a coordinates-to-spikes converter. This converter takes coordinate values in [0, 1] and maps them to spike times in interval [t_early, t_late]. A spike is indicated by a 1 on a dense time axis of length seq_length with temporal resolution dt. Further, it adds a bias spike at time t_bias, if t_bias is not None.

Parameters

• seq_length – Number of time steps in the resulting time sequence. The effective time length is given by seq_length * dt.

• t_early – The earliest time a spike will occur.

• t_late – The latest time a spike will occur.

• dt – The temporal resolution.

• t_bias – The time a bias spike occurs.

Methods

__init__(seq_length, t_early, t_late[, dt, …])	Construct a coordinates-to-spikes converter.
forward(coordinate_values)	Convert coordinate values of to dense spike tensor of zeros and ones.

Attributes

forward(coordinate_values: torch.Tensor) → torch.Tensor

Convert coordinate values of to dense spike tensor of zeros and ones.

Parameters

coordinate_values – Tensor with values in [0, 1], shaped (batch_size, num_channels)

Returns

Returns a dense tensor of shape (batch_size, seq_length, num_channels)

training: bool