Signature
np.linalg.norm(x, ord=None, axis=None)What it does
np.linalg.norm computes vector and matrix norms. Row-wise L2 norms are the usual way to build distance features, normalize embeddings, or check how far a prediction moved from a baseline.
Parameters
xarray_like- Vector or matrix to measure.
ordint or str, optionaloptional- Norm order; default is the L2 norm for vectors.
axisint or tuple, optionaloptional- Axis or axes to reduce.
Returns
ndarray or float
Example
pythonnp.linalg.norm example
import numpy as np
vec = np.array([[3.0, 4.0], [5.0, 12.0]])
print(np.linalg.norm(vec, axis=1))Output
[ 5. 13.]
Watch out for
- A huge norm can mean a real outlier or an unscaled feature; check units.
- For squared distances skip the square root and save a little work.