wnet

library0.9.16pypypi✓ verified 87d ago

Tools for calculating the Wasserstein metric between probability distributions using a network flow algorithm. Version 0.9.16, active development with occasional releases.

analytics

pip install wnet

github ↗homepage ↗

INSTALL

IMPORT

SIG · WNET

wnet

analyticspythonv0.9.16

harness data pending

Install & Compatibility

Where this runs

No compatibility data collected yet for this library.

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

WassersteinNetwork

✓ from wnet import WassersteinNetwork

primary class for computation

L2Cost

✓ from wnet.costs import L2Cost

✗ import L2Cost from wnet

cost functions are in submodule

EMDGauss

✓ from wnet.emd import EMDGauss

Gaussian EMD solver

Quickstart for computing Wasserstein distance between two point sets.

import numpy as np from wnet import WassersteinNetwork from wnet.costs import L2Cost np.random.seed(42) P = np.random.rand(100, 2) Q = np.random.rand(100, 2) cost = L2Cost() net = WassersteinNetwork(P, Q, cost) res = net.solve() print(res.wasserstein_distance)

Debug

Known issues

gotchaCost function must be instantiated before passing to WassersteinNetwork; passing a class instead of an instance leads to cryptic errors.

fix

Always create cost = L2Cost() then use WassersteinNetwork(P, Q, cost).

affects: all

gotchaThe library uses network simplex algorithm which can be slow for large datasets (>10k points). Consider downsampling or using entropic regularized methods if speed is critical.

fix

Downsample input data or use alternative libraries (e.g., POT) for large-scale optimal transport.

affects: all

gotchaWassersteinNetwork modifies inputs in-place? The solver may mutate input arrays; keep original copies if needed.

fix

Pass copies of your data: WassersteinNetwork(P.copy(), Q.copy(), cost).

affects: all

Errors

Common errors & fixes

TypeError: __init__() missing 1 required positional argument: 'cost'

Cost function not provided when creating WassersteinNetwork.

fix

Ensure you pass a cost instance: net = WassersteinNetwork(P, Q, L2Cost())

ModuleNotFoundError: No module named 'wnet.costs'

Older version or incomplete installation. Costs submodule is part of wnet.

fix

Reinstall with pip install --upgrade wnet and import using from wnet.costs import L2Cost

Upgrade

Version history

0.9.16latest on PyPI · released May 20, 2026

Audit

Dependencies

numpyrequiredarray operations

scipyrequiredsparse matrix support

Agent activity

33 hits · last 30 days

node

Anthropic

OpenAI (training)

Resources

githubhttps://github.com/michalsta/wnet.git ↗

homepagehttps://github.com/michalsta/wnet.git ↗