ezmsg.event.kernel_insert

Insert kernels at sparse event locations to produce dense signals.

This module provides efficient sparse-to-dense conversion by inserting kernel waveforms at event locations. Overlapping kernels are summed.

Classes

class SparseKernelInserter(*args, **kwargs)

Bases: BaseStatefulTransformer[SparseKernelInserterSettings, AxisArray, AxisArray, SparseKernelInserterState]

Insert kernels at sparse event locations, producing dense output.

Input: AxisArray with sparse.COO data where:

• coords[0]: sample indices (time)

• coords[1]: channel indices

• data: event values (used for MultiKernel selection or scaling)

Output: AxisArray with dense data containing inserted kernels.

Features:

• Handles chunk boundaries seamlessly (kernel tails carry over)

• Overlapping kernels are summed additively

• Supports acausal kernels (pre_samples > 0)

• Efficient O(n_events * kernel_length) instead of dense convolution

class SparseKernelInserterSettings(kernel=None, scale_by_value=False, output_dtype=<class 'numpy.float64'>)

Bases: Settings

Settings for SparseKernelInserter.

Parameters:

• kernel (Kernel | MultiKernel | None)

• scale_by_value (bool)

• output_dtype (type[Any] | dtype[Any] | _SupportsDType[dtype[Any]] | tuple[Any, Any] | list[Any] | _DTypeDict | str | None)

kernel: Kernel | MultiKernel | None = None

Kernel to insert at event locations. - Kernel: Same kernel for all events. - MultiKernel: Different kernels based on event value. - None: Events are treated as unit impulses (delta functions).

scale_by_value: bool = False

If True, scale kernel amplitude by event value. If False, event value is used only for MultiKernel selection.

output_dtype

Data type for output array.

alias of float64

__init__(kernel=None, scale_by_value=False, output_dtype=<class 'numpy.float64'>)

Parameters:

• kernel (Kernel | MultiKernel | None)

• scale_by_value (bool)

• output_dtype (type[Any] | dtype[Any] | _SupportsDType[dtype[Any]] | tuple[Any, Any] | list[Any] | _DTypeDict | str | None)

Return type:

None

class SparseKernelInserterState

Bases: object

State for SparseKernelInserter.

pending: ndarray[tuple[Any, ...], dtype[floating]] | None = None

pending_length: int = 0

class SparseKernelInserterUnit(*args, settings=None, **kwargs)

Bases: BaseTransformerUnit[SparseKernelInserterSettings, AxisArray, AxisArray, SparseKernelInserter]

Unit for SparseKernelInserter.

Parameters:

settings (Settings | None)

SETTINGS

alias of SparseKernelInserterSettings