SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1903.04154 · 2019

Fisher-Bures Adversary Graph Convolutional Networks

arXiv · PDF · Open in the Atlas

Code that ran

6 of 8 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; 0 honoured the contract we drafted, 0 violated it and 5 have no contract verdict; 1 failure of our instrument; 2 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/1903.04154.json · badge: /api/ran/1903.04154.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
stellargraph/FisherGCN canonical (classifier label)
classifier basis: census
6 ran · 2 no recorded run
FunctionStatusWhere it lives
construct_feed_dict Ran
ran on a synthesized input
stellargraph/FisherGCN/gcn/utils.py
code served (permissive licence) · get_code("077e103b082dc6b3")
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 dict. 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 numpy as np

torch.manual_seed(0)

# Create a simple placeholder class that mimics TensorFlow placeholder behavior
class Placeholder:
    def __init__(self, name):
        self.name = name
        self.indices = f"{name}_indices"
        self.values = f"{name}_values"

# Construct features as a tuple of (indices, values) for sparse representation
features_indices = torch.tensor([[0, 1], [1, 2], [0, 3]], dtype=torch.long)
features_values = torch.tensor([1.0, 2.0, 3.0], dtype=torch.float32)
features = (features_indices, features_values)

# Construct support as a list of adjacency matrices
support = [
    torch.tensor([[1.0, 0.5], [0.5, 1.0]], dtype=torch.float32),
    torch.tensor([[0.8, 0.2], [0.2, 0.8]], dtype=torch.float32)
]

# Construct labels as a tensor (2 samples, 4 classes)
labels = torch.tensor([[1, 0, 0, 0], [0, 1, 0, 0]], dtype=torch.float32)

# Construct labels_mask (2 samples, binary mask)
labels_mask = torch.tensor([True, True], dtype=torch.bool)

# Construct placeholders dict
placeholders = {
    'features': Placeholder('features'),
    'support': [Placeholder(f'support_{i}') for i in range(len(support))],
    'labels': 'labels_placeholder',
    'labels_mask': 'labels_mask_placeholder',
    'noise': 'noise_placeholder',
    'dropout': 'dropout_placeholder'
}

# Construct noise as a scalar or small tensor
noise = 0.1

# Construct dropout as a scalar
dropout = 0.5

ARGS = {
    'features': features,
    'support': support,
    'labels': labels,
    'labels_mask': labels_mask,
    'placeholders': placeholders,
    'noise': noise,
    'dropout': dropout
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
create_subgraph Ran
ran on a synthesized input
stellargraph/FisherGCN/gcn/data_io.py
code served (permissive licence) · get_code("05a963d9efb91665")
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 SparseGraph. 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
import scipy.sparse as sp

class SparseGraph:
    """Minimal SparseGraph implementation matching the function's usage."""
    def __init__(self, adj_matrix, attr_matrix=None, labels=None, node_names=None):
        self.adj_matrix = adj_matrix
        self.attr_matrix = attr_matrix
        self.labels = labels
        self.node_names = node_names
    
    def num_nodes(self):
        return self.adj_matrix.shape[0]

# Create a small sparse adjacency matrix (5x5)
adj_data = np.array([0, 1, 2, 3, 4, 1, 0, 2, 3, 4])
adj_row = np.array([0, 0, 1, 1, 2, 2, 3, 3, 4, 4])
adj_col = np.array([0, 1, 0, 2, 1, 3, 2, 4, 3, 4])
adj_matrix = sp.csr_matrix((adj_data, (adj_row, adj_col)), shape=(5, 5))

# Create attribute matrix (5 nodes x 4 features)
attr_matrix = np.random.randn(5, 4).astype(np.float32)

# Create labels (5 nodes)
labels = np.array([0, 1, 0, 1, 0])

# Create node names
node_names = np.array(['node_0', 'node_1', 'node_2', 'node_3', 'node_4'])

# Create SparseGraph instance
sparse_graph = SparseGraph(
    adj_matrix=adj_matrix,
    attr_matrix=attr_matrix,
    labels=labels,
    node_names=node_names
)

ARGS = {
    'sparse_graph': sparse_graph,
    '_sentinel': None,
    'nodes_to_remove': None,
    'nodes_to_keep': [0, 2, 4]
}
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 SparseGraph — a SparseGraph output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SparseGraph output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
eliminate_self_loops Ran
ran on a synthesized input
stellargraph/FisherGCN/gcn/data_io.py
code served (permissive licence) · get_code("59d0563fb1fca77c")
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 SimpleGraph. 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 scipy.sparse as sp
import numpy as np

class SimpleGraph:
    def __init__(self, adj_matrix):
        self.adj_matrix = adj_matrix

# Create a small sparse adjacency matrix with self-loops
# 4x4 matrix with some self-loops
rows = [0, 0, 1, 1, 2, 2, 3, 3]
cols = [0, 1, 1, 2, 2, 3, 3, 0]
data = [1, 1, 1, 1, 1, 1, 1, 1]
adj_matrix = sp.csr_matrix((data, (rows, cols)), shape=(4, 4))

G = SimpleGraph(adj_matrix)

ARGS = {
    "G": G
}
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 SimpleGraph — a SimpleGraph output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SimpleGraph output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
largest_connected_components Ran
ran on a synthesized input
stellargraph/FisherGCN/gcn/data_io.py
code served (permissive licence) · get_code("88a77fd6e7ad9f83")
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 SparseGraph. 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 scipy.sparse as sp
import numpy as np

class SparseGraph:
    """Minimal SparseGraph class with required attributes and methods."""
    def __init__(self, adj_matrix, attr_matrix=None, labels=None, node_names=None):
        self.adj_matrix = adj_matrix
        self.attr_matrix = attr_matrix
        self.labels = labels
        self.node_names = node_names
    
    def num_nodes(self):
        return self.adj_matrix.shape[0]

# Create a small sparse adjacency matrix for a connected graph
# 5 nodes with some connections
row = np.array([0, 1, 1, 2, 2, 3, 4])
col = np.array([1, 0, 2, 1, 3, 2, 0])
data = np.array([1, 1, 1, 1, 1, 1, 1])
adj_matrix = sp.csr_matrix((data, (row, col)), shape=(5, 5))

# Create optional attribute matrix (5 nodes, 4 features)
attr_matrix = np.random.randn(5, 4)

# Create optional labels (5 nodes)
labels = np.array([0, 1, 0, 1, 0])

# Create optional node names
node_names = np.array(['node_0', 'node_1', 'node_2', 'node_3', 'node_4'])

# Create the SparseGraph instance
sparse_graph = SparseGraph(
    adj_matrix=adj_matrix,
    attr_matrix=attr_matrix,
    labels=labels,
    node_names=node_names
)

ARGS = {
    'sparse_graph': sparse_graph,
    'n_components': 1
}
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 SparseGraph — a SparseGraph output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SparseGraph output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
masked_accuracy Ran
ran on a synthesized input
stellargraph/FisherGCN/gcn/metrics.py
code served (permissive licence) · get_code("7db4f095bc7a6cbe")
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 EagerTensor, 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 tensorflow as tf
import numpy as np

# Create small sample tensors for the masked_accuracy function
batch_size = 2
num_classes = 4

# preds: logits with shape (batch_size, num_classes)
# These will be passed to argmax to get predicted class
preds = tf.constant([
    [0.1, 0.2, 0.5, 0.2],
    [0.3, 0.1, 0.1, 0.5]
], dtype=tf.float32)

# labels: one-hot encoded labels with shape (batch_size, num_classes)
# These will be passed to argmax to get true class
labels = tf.constant([
    [0.0, 0.0, 1.0, 0.0],
    [0.0, 0.0, 0.0, 1.0]
], dtype=tf.float32)

# mask: binary mask with shape (batch_size,)
# Used to mask out certain examples in the accuracy calculation
mask = tf.constant([1.0, 1.0], dtype=tf.float32)

ARGS = {
    "preds": preds,
    "labels": labels,
    "mask": mask
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, 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
sparse_to_tuple Ran
ran; our contract draft was wrong, not the code
stellargraph/FisherGCN/gcn/utils.py
code served (permissive licence) · get_code("eaa82e9c7a22e930")
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
import scipy.sparse as sp

# Create a small sparse matrix
row = np.array([0, 1, 2, 3])
col = np.array([1, 2, 0, 3])
data = np.array([1.0, 2.0, 3.0, 4.0])
sparse_matrix = sp.coo_matrix((data, (row, col)), shape=(4, 4))

# Test with single sparse matrix
ARGS = {
    "sparse_mx": sparse_matrix
}

# Also test with a list of sparse matrices to verify the list branch
sparse_matrix_2 = sp.csr_matrix(([5.0, 6.0], ([0, 1], [1, 2])), shape=(3, 4))
sparse_matrix_3 = sp.lil_matrix((2, 5))
sparse_matrix_3[0, 1] = 7.0
sparse_matrix_3[1, 3] = 8.0

ARGS_list = {
    "sparse_mx": [sparse_matrix, sparse_matrix_2, sparse_matrix_3]
}

# Use the single matrix version as default
ARGS = {
    "sparse_mx": sparse_matrix
}
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
block_krylov Not yet run
unverified (harvested, no recorded run)
stellargraph/FisherGCN/gcn/block_krylov.py
code served (permissive licence) · get_code("e4f3aae94a04a97a")
How it ran

no invocation recorded.

masked_softmax_cross_entropy Not yet run
unverified (harvested, no recorded run)
stellargraph/FisherGCN/gcn/metrics.py
code served (permissive licence) · get_code("38c4e4b6d2de6143")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 7 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
sparse_to_tuple 7 other papers
7 later
aaai_20388 (2022 (year only))
RepBin: Constraint-Based Graph Representation Learning for Metagenomic Binning
ijcai2022_0310 (2022 (year only))
Beyond Homophily: Structure-aware Path Aggregation Graph Neural Network
2203.03315 (2022-03)
Deep Reinforcement Learning for Entity Alignment
and 4 more not named here

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

In a graph convolutional network, we assume that the graph $G$ is generated wrt some observation noise. During learning, we make small random perturbations $ΔG$ of the graph and try to improve generalization. Based on quantum information geometry, $ΔG$ can be characterized by the eigendecomposition of the graph Laplacian matrix. We try to minimize the loss wrt the perturbed $G+Δ{G}$ while making $Δ{G}$ to be effective in terms of the Fisher information of the neural network. Our proposed model can consistently improve graph convolutional networks on semi-supervised node classification tasks with reasonable computational overhead. We present three different geometries on the manifold of graphs: the intrinsic geometry measures the information theoretic dynamics of a graph; the extrinsic geometry characterizes how such dynamics can affect externally a graph neural network; the embedding geometry is for measuring node embeddings. These new analytical tools are useful in developing a good understanding of graph neural networks and fostering new techniques.

For agents

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

get_harvested_code_for_paper("1903.04154")
get_code_for_paper("1903.04154")
have("1903.04154")

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

curl https://syntology.ai/api/ran/1903.04154.json

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

[![Syntology run record](https://syntology.ai/api/ran/1903.04154.svg)](https://syntology.ai/paper/1903.04154)

Connect an agent — have() is free.