Normal — Random Normal Weight Initializer¶
Samples weights from a normal (Gaussian) distribution with configurable mean and standard deviation.
Overview¶
Normal draws weight values from a normal distribution centered at mean with standard deviation stddev. It provides a flexible, distribution-based initialization strategy for neural network weights. Reproducible via seed.
Key Features¶
- Configurable mean and standard deviation
- Reproducible via
seed - Returns
np.float64array by default - Alias:
RandomNormal
Parameters¶
| Parameter | Type | Default | Description |
|---|---|---|---|
mean |
float |
0.0 |
Mean of the normal distribution |
stddev |
float |
0.05 |
Standard deviation of the normal distribution |
seed |
int | None |
None |
Random seed for reproducibility |
dtype |
np.dtype |
np.float64 |
Output data type |
Methods¶
__call__(shape)¶
Return normal random samples with the requested shape.
Example Usage¶
from machinegnostics.magnet import Normal
init = Normal(mean=0.0, stddev=0.05, seed=42)
W = init((10, 20))
print(W.shape) # (10, 20)
Author: Nirmal Parmar