Gaussian¶

Gaussian(params)

Creates an n-dimensional Gaussian kernel. Gaussian kernels are used by GaussInput, Convolution, and NeuralField.

Parameters¶

Parameter	Description
`sigma`	Tuple of standard deviations, one per dimension.
`amplitude`	Kernel amplitude. Negative amplitudes create inhibitory kernels.
`normalized`	If `True`, normalize each component before applying amplitude.
`shape`	Optional explicit kernel shape. If omitted, a size is estimated from `sigma`.
`center`	Optional center index. Defaults to the center of `shape`.
`max_shape`	Optional maximum shape. Oversized kernels are cropped.
`factorized`	If `True`, store separable one-dimensional factors. Defaults to `True`.

Example¶

import juniper as jp

kernel = jp.Gaussian({
    "shape": (50,),
    "sigma": (3,),
    "amplitude": 5.0,
    "normalized": True,
    "factorized": True,
})

Use factorized=False when a full materialized kernel is required.