Skip to content

Uniform — Uniform Random Weight Initializer

Samples weights from a uniform distribution over [-limit, +limit], where limit is derived from the input dimension.


Overview

Uniform draws weight values from a uniform distribution centered at zero, scaled by sqrt(1 / fan_in). It provides a simple, distribution-free initialization strategy that works reasonably well across a variety of activation functions. Reproducible via seed.


Key Features

  • Uniform sampling over [-limit, +limit] with limit = sqrt(1 / fan_in)
  • Works as a general-purpose default initializer
  • Reproducible via seed
  • Returns np.float64 array by default

Parameters

Parameter Type Default Description
seed int | None None Random seed for reproducibility
dtype np.dtype np.float64 Output data type

Methods

__call__(shape)

Return uniform random samples scaled by the fan-in of the target shape.


Example Usage

from machinegnostics.magnet import Uniform
init = Uniform(seed=42)
W = init((10, 20))
print(W.shape)  # (10, 20)
print(W.min(), W.max())  # bounds around 0

Author: Nirmal Parmar