SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1902.02384 · 2019

Global Explanations of Neural Networks: Mapping the Landscape of Predictions

arXiv · PDF · Open in the Atlas

Code that ran

8 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 8 have no contract verdict; 0 failures of our instrument; 0 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/1902.02384.json · badge: /api/ran/1902.02384.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
capitalone/global-attribution-mapping canonical (classifier label)
classifier basis: census
8 ran · 0 no recorded run
FunctionStatusWhere it lives
finalize Ran
ran on a synthesized input
capitalone/global-attribution-mapping/gam/clustering.py
code served (permissive licence) · get_code("254c99c194d53ccc")
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, ndim 1, dtype family float; 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 torch
import numpy as np

torch.manual_seed(0)

# The function expects existingAggregate to be a tuple of (count, mean, m2)
# where:
# - count is an integer representing the number of observations
# - mean is a numeric value (the running mean)
# - m2 is a numeric value (sum of squared differences)

count = 5
mean = 2.5
m2 = 10.0

existingAggregate = (count, mean, m2)

ARGS = {
    "existingAggregate": existingAggregate
}
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 tuple, ndim 1, dtype family float — a tuple 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
ktau_weighted_distance Ran
ran on a synthesized input
capitalone/global-attribution-mapping/gam/kendall_tau_distance.py
code served (permissive licence) · get_code("52d5d1c4dfa22c28")
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 int, ndim 0, dtype family int, 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 numpy as np

# Construct arguments for ktau_weighted_distance
# The function takes two lists of weighted rankings
# Each ranking should be a list of positive numbers that sum to 1

# Create first weighted ranking - 3 items with weights summing to 1
r_1 = [0.1, 0.2, 0.7]

# Create second weighted ranking - 3 items with weights summing to 1
r_2 = [0.15, 0.25, 0.6]

ARGS = {
    "r_1": r_1,
    "r_2": r_2
}
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 int, ndim 0, dtype family int, 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
mergeSortInversions Ran
ran on a synthesized input
capitalone/global-attribution-mapping/gam/kendall_tau_distance.py
code served (permissive licence) · get_code("be812738cec9b955")
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

# Construct valid arguments for mergeSortInversions
# The function expects:
# - arr: 1D array of tuples (y_i, ind) where y_i is a numeric value and ind is an index
# - indList: list of pairs of indices (starts empty, gets populated during recursion)

# Create a small 1D array of tuples
arr = [
    (5.0, 0),
    (2.0, 1),
    (8.0, 2),
    (1.0, 3),
]

# Start with empty indList (gets populated during the merge sort)
indList = []

ARGS = {
    "arr": arr,
    "indList": indList,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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
prep_data Ran
ran on a synthesized input
capitalone/global-attribution-mapping/gam/kendall_tau_distance.py
code served (permissive licence) · get_code("7e2cbb7b41ee832a")
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 2, 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 torch
import numpy as np

torch.manual_seed(0)

# X and Y are 1D arrays of weighted rankings
# The function zips them together with indices, sorts by X (then Y for ties),
# and removes X from the result
X = np.array([3.0, 1.0, 2.0, 1.5])
Y = np.array([10, 20, 30, 40])

ARGS = {
    "X": X,
    "Y": Y
}
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 list, ndim 2, 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
search_singles Ran
ran on a synthesized input
capitalone/global-attribution-mapping/gam/clustering.py
code served (permissive licence) · get_code("b1dcbd1d77f9c7c1")
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.spatial.distance import cdist

# Set seed for reproducibility
np.random.seed(42)

# Create a small sample dataset X with shape (5, 3)
# 5 data points, 3 features each
X = np.array([
    [1.0, 2.0, 3.0],
    [4.0, 5.0, 6.0],
    [7.0, 8.0, 9.0],
    [2.0, 3.0, 4.0],
    [5.0, 6.0, 7.0]
], dtype=np.float64)

# solution_ids: list of indices to consider as potential medoids
solution_ids = [0, 1, 2]

# dist_func: a distance metric supported by scipy.spatial.distance.cdist
dist_func = 'euclidean'

# d_nearest: array of nearest distances for each point
# Shape should match X (number of points,)
d_nearest = np.array([0.5, 1.0, 1.5, 0.8, 1.2], dtype=np.float64)

ARGS = {
    'X': X,
    'solution_ids': solution_ids,
    'dist_func': dist_func,
    'd_nearest': d_nearest
}
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
spearman_squared_distance Ran
ran on a synthesized input
capitalone/global-attribution-mapping/gam/spearman_distance.py
code served (permissive licence) · get_code("870ee7531c26e47c")
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 float, 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

# Create two 1D numpy arrays of normalized local attributions
# The arrays should sum to approximately 1 to match the docstring description
a = np.array([0.1, 0.2, 0.3, 0.4], dtype=np.float64)
b = np.array([0.15, 0.25, 0.35, 0.25], dtype=np.float64)

ARGS = {
    "a": a,
    "b": b,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, 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
spearman_squared_distance_legacy Ran
ran on a synthesized input
capitalone/global-attribution-mapping/gam/spearman_distance.py
code served (permissive licence) · get_code("a067b5ff8113cf94")
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 float, 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 numpy as np

torch.manual_seed(0)

# Based on the function signature and body:
# - r_1 and r_2 are lists of weighted rankings
# - Each entry should be positive and sum to 1
# - The function iterates through them with zip and performs arithmetic

# Create two valid weighted ranking lists
r_1 = [0.1, 0.2, 0.3, 0.4]  # sums to 1.0, all positive
r_2 = [0.25, 0.25, 0.25, 0.25]  # sums to 1.0, all positive

ARGS = {
    "r_1": r_1,
    "r_2": r_2,
}
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 float, 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
update Ran
ran on a synthesized input
capitalone/global-attribution-mapping/gam/clustering.py
code served (permissive licence) · get_code("1d998fd0012807db")
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

# Construct arguments for the update function
# The function expects:
# - existingAggregate: a tuple of (count, mean, m2) representing Welford's algorithm state
# - new_values: numpy array of new values to incorporate

# Initialize the existing aggregate state
count = 5
mean = np.array([1.0, 2.0, 3.0, 1.5])
m2 = np.array([0.5, 0.8, 1.2, 0.3])
existingAggregate = (count, mean, m2)

# Create new values as a numpy array (batch of values)
new_values = np.array([
    [1.2, 2.1, 3.0, 1.6],
    [0.9, 1.9, 2.9, 1.4]
])

ARGS = {
    "existingAggregate": existingAggregate,
    "new_values": new_values
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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

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, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 8 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.

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

A barrier to the wider adoption of neural networks is their lack of interpretability. While local explanation methods exist for one prediction, most global attributions still reduce neural network decisions to a single set of features. In response, we present an approach for generating global attributions called GAM, which explains the landscape of neural network predictions across subpopulations. GAM augments global explanations with the proportion of samples that each attribution best explains and specifies which samples are described by each attribution. Global explanations also have tunable granularity to detect more or fewer subpopulations. We demonstrate that GAM's global explanations 1) yield the known feature importances of simulated data, 2) match feature weights of interpretable statistical models on real data, and 3) are intuitive to practitioners through user studies. With more transparent predictions, GAM can help ensure neural network decisions are generated for the right reasons.

For agents

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

get_harvested_code_for_paper("1902.02384")
get_code_for_paper("1902.02384")
have("1902.02384")

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

curl https://syntology.ai/api/ran/1902.02384.json

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

[![Syntology run record](https://syntology.ai/api/ran/1902.02384.svg)](https://syntology.ai/paper/1902.02384)

Connect an agent — have() is free.