SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2106.01954 · 2021

Do Neural Optimal Transport Solvers Work? A Continuous Wasserstein-2 Benchmark

arXiv · PDF · Open in the Atlas

Code that ran

11 of 19 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 3 honoured the contract we drafted, 0 violated it and 5 have no contract verdict; 3 failures of our instrument; 8 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2106.01954.json · badge: /api/ran/2106.01954.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
iamalexkorotin/Wasserstein2GenerativeNetworks canonical (classifier label)
classifier basis: census
4 ran · 3 no recorded run
iamalexkorotin/Wasserstein2Benchmark canonical (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
iamalexkorotin/Wasserstein2Barycenters community (classifier label)
classifier basis: plan_row
5 ran · 4 no recorded run
copy not recorded community
not in the paper's declared repositories
0 ran · 1 no recorded run
FunctionStatusWhere it lives
compute_l1_norm Ran
ran on a synthesized input
iamalexkorotin/Wasserstein2GenerativeNetworks/src/tools.py
code served (permissive licence) · get_code("8c031ff08115e9a3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.linear2(x)
        return x

model = SimpleModel()

ARGS = {
    "model": model
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
ewma Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
iamalexkorotin/Wasserstein2Barycenters/src/tools.py
code served (permissive licence) · get_code("07c1c7fb63590df6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned ndarray, ndim 1, dtype family float, shape (10,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

x = np.array([1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0])
span = 200

ARGS = {
    'x': x,
    'span': span
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (10,) — shape (10,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_barycenter_cov Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
iamalexkorotin/Wasserstein2Barycenters/src/benchmarks.py
code served (permissive licence) · get_code("ed9cd1d0db6863a4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (4, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

np.random.seed(0)

# Create small positive definite covariance matrices
dim = 4
num_covs = 3

# Generate random positive definite matrices using eigendecomposition
covs = []
for i in range(num_covs):
    # Create random matrix
    A = np.random.randn(dim, dim).astype(np.float32)
    # Make it positive definite by A @ A.T
    cov = (A @ A.T).astype(np.float32)
    covs.append(cov)

# Create weights that sum to 1
alphas = np.array([0.2, 0.3, 0.5], dtype=np.float32)

ARGS = {
    "covs": covs,
    "alphas": alphas,
    "max_iter": 100,
    "tol": 1e-08,
    "verbose": False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (4, 4) — shape (4, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_linear_transport Ran
ran, honoured the contract we drafted
iamalexkorotin/Wasserstein2Barycenters/src/benchmarks.py
code served (permissive licence) · get_code("101f00b70e1555e6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
from scipy.linalg import sqrtm

np.random.seed(0)

# Create small positive definite covariance matrices
# Using dimension 4 for consistency
dim = 4

# Generate random positive definite matrices by A @ A.T
A1 = np.random.randn(dim, dim)
cov1 = A1 @ A1.T

A2 = np.random.randn(dim, dim)
cov2 = A2 @ A2.T

# Create mean vectors
mean1 = np.random.randn(dim)
mean2 = np.random.randn(dim)

ARGS = {
    "mean1": mean1,
    "cov1": cov1,
    "mean2": mean2,
    "cov2": cov2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_linear_transport_cost Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
iamalexkorotin/Wasserstein2Benchmark/src/map_benchmark.py
code served (permissive licence) · get_code("b431a14584153806")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned float64, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

np.random.seed(0)

# Create two Gaussian distributions with different means and covariances
# mu1, mu2 are mean vectors (1D arrays)
# sigma1, sigma2 are covariance matrices (2D arrays, symmetric positive definite)

dim = 5

# Mean vectors
mu1 = np.random.randn(dim)
mu2 = np.random.randn(dim)

# Create valid covariance matrices (symmetric positive definite)
# Method: A @ A.T is always positive semi-definite
A1 = np.random.randn(dim, dim)
sigma1 = A1 @ A1.T + np.eye(dim) * 0.1  # Add small diagonal for numerical stability

A2 = np.random.randn(dim, dim)
sigma2 = A2 @ A2.T + np.eye(dim) * 0.1

eps = 1e-06

ARGS = {
    'mu1': mu1,
    'sigma1': sigma1,
    'mu2': mu2,
    'sigma2': sigma2,
    'eps': eps
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_linear_transport_map Ran
ran, honoured the contract we drafted
iamalexkorotin/Wasserstein2Benchmark/src/map_benchmark.py
code served (permissive licence) · get_code("77a46a75384b972e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

np.random.seed(0)

# Create two 3-dimensional Gaussian distributions
dim = 3

# First distribution
mean1 = np.random.randn(dim)
# Create a valid covariance matrix (positive definite)
A1 = np.random.randn(dim, dim)
cov1 = A1 @ A1.T + np.eye(dim) * 0.1

# Second distribution
mean2 = np.random.randn(dim)
# Create another valid covariance matrix (positive definite)
A2 = np.random.randn(dim, dim)
cov2 = A2 @ A2.T + np.eye(dim) * 0.1

ARGS = {
    'mean1': mean1,
    'cov1': cov1,
    'mean2': mean2,
    'cov2': cov2
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
imread Ran
ran, honoured the contract we drafted
iamalexkorotin/Wasserstein2GenerativeNetworks/src/fid_score.py
pointer only (licence: MIT); the stored copy get_code serves, from zmtomorrow/mmdgs_neurips, is pointer only (licence: NONE) · get_code("aa4cae054b657ba2")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from zmtomorrow/mmdgs_neurips, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

plot_latent_pca Ran
ran on a synthesized input
iamalexkorotin/Wasserstein2GenerativeNetworks/src/plotters.py
code served (permissive licence) · get_code("28966a12560fcf99")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import numpy as np
from sklearn.decomposition import PCA

torch.manual_seed(0)
np.random.seed(0)

# Create a small decoder module that implements the push method
class SimpleDecoder(nn.Module):
    def __init__(self, input_dim, output_dim):
        super().__init__()
        self.linear = nn.Linear(input_dim, output_dim)
    
    def forward(self, x):
        return self.linear(x)
    
    def push(self, x):
        """Transform input through the decoder"""
        return self.forward(x)

# Small tensors: batch_size=2, latent_dim=4
batch_size = 2
latent_dim = 4
output_dim = 4

# Create Z (initial latent codes)
Z = torch.randn(batch_size, latent_dim)

# Create E (latent space embeddings)
E = torch.randn(batch_size, latent_dim)

# Create decoder
D = SimpleDecoder(latent_dim, output_dim)

# Create PCA model fitted on the data
Z_flat = Z.cpu().detach().numpy().reshape(batch_size, -1)
E_flat = E.cpu().detach().numpy().reshape(batch_size, -1)
combined_data = np.vstack([Z_flat, E_flat])

pca = PCA(n_components=min(4, combined_data.shape[1]))
pca.fit(combined_data)

# Arguments dictionary
ARGS = {
    'Z': Z,
    'E': E,
    'D': D,
    'pca': pca,
    'n_pairs': 2,
    'show': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, numpy, sklearn, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
plot_training_phase Ran
ran on a synthesized input
iamalexkorotin/Wasserstein2Barycenters/src/plotters.py
code served (permissive licence) · get_code("32c405ba8da2078b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import numpy as np
from sklearn.decomposition import PCA

torch.manual_seed(0)
np.random.seed(0)

# Create a simple sampler class
class SimpleSampler:
    def __init__(self, dim=4, batch_size=2):
        self.dim = dim
        self.batch_size = batch_size
    
    def sample(self, batch_size):
        return torch.randn(batch_size, self.dim)

# Create a simple push operator (identity-like)
class SimplePushOperator:
    def __init__(self, scale=1.0):
        self.scale = scale
    
    def push(self, X):
        return X * self.scale

# Create a simple generator
class SimpleGenerator(nn.Module):
    def __init__(self, latent_dim=3, output_dim=4):
        super().__init__()
        self.linear = nn.Linear(latent_dim, output_dim)
    
    def forward(self, Z):
        return self.linear(Z)

# Create benchmark object
class SimpleBenchmark:
    def __init__(self, num=2):
        self.num = num
        self.samplers = [SimpleSampler(dim=4) for _ in range(num)]
        self.alphas = [1.0 / num for _ in range(num)]
        self.bar_sampler = SimpleSampler(dim=4)
        self.gauss_bar_sampler = None

# Create PCA object
X_train = torch.randn(100, 4).numpy()
pca = PCA(n_components=2)
pca.fit(X_train)

# Create arguments
benchmark = SimpleBenchmark(num=2)
D_list = [SimplePushOperator(scale=1.0) for _ in range(2)]
D_conj_list = [SimplePushOperator(scale=1.1) for _ in range(2)]
G = SimpleGenerator(latent_dim=3, output_dim=4)
Z_sampler = SimpleSampler(dim=3)

ARGS = {
    'benchmark': benchmark,
    'pca': pca,
    'D_list': D_list,
    'D_conj_list': D_conj_list,
    'G': G,
    'Z_sampler': Z_sampler,
    'plot_batchsize': 2,
    'partsize': (5, 5),
    'dpi': 100
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, numpy, sklearn, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
plot_training_phase_palettes Ran
ran on a synthesized input
iamalexkorotin/Wasserstein2Barycenters/src/plotters.py
code served (permissive licence) · get_code("d19df8e62477a487")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
import matplotlib.pyplot as plt
from torch import nn

torch.manual_seed(0)

class SimpleSampler:
    """Sampler that generates random RGB points"""
    def sample(self, batch_size):
        return torch.rand(batch_size, 3)

class SimpleTransform(nn.Module):
    """Simple transform that applies a linear map to RGB points"""
    def __init__(self, in_dim=3, out_dim=3):
        super().__init__()
        self.linear = nn.Linear(in_dim, out_dim)
    
    def push(self, X):
        # Apply linear transform and clip to valid RGB range
        output = self.linear(X)
        return torch.clamp(output, 0, 1)

class Benchmark:
    """Benchmark class with samplers and number of distributions"""
    def __init__(self, num_distributions=2):
        self.num = num_distributions
        self.samplers = [SimpleSampler() for _ in range(num_distributions)]

# Create benchmark with 2 distributions
benchmark = Benchmark(num_distributions=2)

# Create simple transforms for pushing and conjugate pushing
D_list = [SimpleTransform() for _ in range(benchmark.num)]
D_conj_list = [SimpleTransform() for _ in range(benchmark.num)]

ARGS = {
    'benchmark': benchmark,
    'D_list': D_list,
    'D_conj_list': D_conj_list,
    'plot_batchsize': 2,
    'partsize': (5, 5),
    'dpi': 200,
    'elev': 0.0,
    'azim': 40
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
read_images_to_ram Ran
ran on a synthesized input
iamalexkorotin/Wasserstein2GenerativeNetworks/src/tools.py
code served (permissive licence) · get_code("c424fce4b2b9d60d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 4, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import sys
import tempfile
from PIL import Image
import numpy as np

# Create a temporary directory with some test images
temp_dir = tempfile.mkdtemp()

# Create a few small test images and save them
for i in range(3):
    img_array = np.random.randint(0, 256, (10, 10, 3), dtype=np.uint8)
    img = Image.fromarray(img_array, 'RGB')
    img.save(os.path.join(temp_dir, f'test_image_{i}.png'))

# Create one file that will fail to read (not a valid image)
with open(os.path.join(temp_dir, 'not_an_image.txt'), 'w') as f:
    f.write('This is not an image')

ARGS = {
    'path': temp_dir,
    'mode': 'RGB',
    'verbose': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 4, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
fig2data Not yet run
unverified (harvested, no recorded run)
iamalexkorotin/Wasserstein2Barycenters/src/tools.py
code served (permissive licence) · get_code("910de59edc0ed1b0")
How it ran

no invocation recorded.

fig2img Not yet run
unverified (harvested, no recorded run)
iamalexkorotin/Wasserstein2Barycenters/src/tools.py
code served (permissive licence) · get_code("5d9a01fa212bd744")
How it ran

no invocation recorded.

get_activations Not yet run
unverified (harvested, no recorded run)
iamalexkorotin/Wasserstein2GenerativeNetworks/src/fid_score.py
pointer only (licence: MIT); the stored copy get_code serves, from guoqiang-zhang-x/BDIA, is pointer only (licence: NONE) · get_code("52f0b0681052dbb6")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from guoqiang-zhang-x/BDIA, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

get_activations_for_dataloader Not yet run
unverified (harvested, no recorded run)
iamalexkorotin/Wasserstein2GenerativeNetworks/src/fid_score.py
code served (permissive licence) · get_code("43cd9ab2a1a6fd6d")
How it ran

no invocation recorded.

plot_combined_generated Not yet run
unverified (harvested, no recorded run)
iamalexkorotin/Wasserstein2GenerativeNetworks/src/plotters.py
code served (permissive licence) · get_code("0dbfd332ac783bb6")
How it ran

no invocation recorded.

plot_conj Not yet run
unverified (harvested, no recorded run)
this paper's copy was not recorded; identical code first harvested from facebookresearch/w2ot
pointer only · get_code("7fb9bfa9b31ec695")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

push_img Not yet run
unverified (harvested, no recorded run)
iamalexkorotin/Wasserstein2Barycenters/src/plotters.py
code served (permissive licence) · get_code("94417c450f1ed113")
How it ran

no invocation recorded.

symmetrize Not yet run
unverified (harvested, no recorded run)
iamalexkorotin/Wasserstein2Barycenters/src/benchmarks.py
code served (permissive licence) · get_code("c1eb471d75746958")
How it ran

no invocation recorded.

Also present in other papers' code

Of 19 samples lifted from this paper's linked repositories, 19 carry a function identity and 0 do not. Of the 19 distinct functions with an identity, 9 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 0 the same month, 9 later) and 10 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
imread 12 other papers
1 dated earlier, 11 later
2009.05475 (2020-09)
Adversarial score matching and improved sampling for image generation
2112.05134 (2021-12)
A Shared Representation for Photorealistic Driving Simulators
2201.12220 (2022-01)
Neural Optimal Transport
and 9 more not named here
get_activations 10 other papers
10 later
2112.05134 (2021-12)
A Shared Representation for Photorealistic Driving Simulators
2201.12220 (2022-01)
Neural Optimal Transport
2205.15269 (2022-05)
Kernel Neural Optimal Transport
and 7 more not named here
fig2data 8 other papers
2 dated earlier, 6 later
1806.09045 (2018-06)
Variational Wasserstein Clustering
2007.05929 (2020-07)
Data-Efficient Reinforcement Learning with Self-Predictive Representations
2201.12220 (2022-01)
Neural Optimal Transport
and 5 more not named here
fig2img 4 other papers
1 dated earlier, 3 later
1805.09190 (2018-05)
Towards the first adversarially robust neural network model on MNIST
2310.01174 (2023-10)
Light Schrödinger Bridge
2405.14449 (2024-05)
Adversarial Schrödinger Bridge Matching
and 1 more not named here
plot_conj 4 other papers
2 dated earlier, 2 later
1902.07197 (2019-02)
2-Wasserstein Approximation via Restricted Convex Potentials with Application to Improved Training for GANs
1909.13082 (2019-09)
Wasserstein-2 Generative Networks
2110.02999 (2021-10)
Generative Modeling with Optimal Transport Maps
and 1 more not named here
ewma 2 other papers
2 later
2201.12220 (2022-01)
Neural Optimal Transport
2205.15269 (2022-05)
Kernel Neural Optimal Transport
symmetrize 2 other papers
2 later
2201.12245 (2022-01)
Wasserstein Iterative Networks for Barycenter Estimation
2306.10161 (2023-06)
Building the Bridge of Schrödinger: A Continuous Entropic Optimal Transport Benchmark
get_barycenter_cov 1 other paper
1 later
2201.12245 (2022-01)
Wasserstein Iterative Networks for Barycenter Estimation
get_linear_transport 1 other paper
1 later
2201.12245 (2022-01)
Wasserstein Iterative Networks for Barycenter Estimation

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Despite the recent popularity of neural network-based solvers for optimal transport (OT), there is no standard quantitative way to evaluate their performance. In this paper, we address this issue for quadratic-cost transport -- specifically, computation of the Wasserstein-2 distance, a commonly-used formulation of optimal transport in machine learning. To overcome the challenge of computing ground truth transport maps between continuous measures needed to assess these solvers, we use input-convex neural networks (ICNN) to construct pairs of measures whose ground truth OT maps can be obtained analytically. This strategy yields pairs of continuous benchmark measures in high-dimensional spaces such as spaces of images. We thoroughly evaluate existing optimal transport solvers using these benchmark measures. Even though these solvers perform well in downstream tasks, many do not faithfully recover optimal transport maps. To investigate the cause of this discrepancy, we further test the solvers in a setting of image generation. Our study reveals crucial limitations of existing solvers and shows that increased OT accuracy does not necessarily correlate to better results downstream.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2106.01954")
get_code_for_paper("2106.01954")
have("2106.01954")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2106.01954.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2106.01954.svg)](https://syntology.ai/paper/2106.01954)

Connect an agent — have() is free.