SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2007.05721 · ICML · 2020

Towards Robust Classification with Deep Generative Forests

Alvaro Correia, Robert Peharz, Cassio De Campos

arXiv · PDF · Open in the Atlas

Code that ran

10 of 25 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 10 have no contract verdict; 0 failures of our instrument; 15 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/2007.05721.json · badge: /api/ran/2007.05721.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
AlCorreia/GeFs canonical (classifier label)
classifier basis: census
10 ran · 15 no recorded run
FunctionStatusWhere it lives
bincount Ran
ran on a synthesized input
AlCorreia/GeFs/gefs/split.py
code served (permissive licence) · get_code("58b7f952cc8ba9bd")
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 1, dtype family int, shape (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
from numba import njit, int64, float64, optional, prange, deferred_type, types, boolean

# Analyze the function:
# - data: used with .size and indexing data[j], so needs to be an array of integers
# - n: used as size argument for np.zeros, so needs to be an integer
# The function counts occurrences of values in data, so data should contain indices
# that are valid (0 to n-1)

# Create a small numpy array with integer values in valid range
data = np.array([0, 1, 2, 1, 0, 3, 2, 1], dtype=np.int64)
n = 4  # number of bins, should be >= max(data) + 1

ARGS = {
    'data': data,
    'n': n
}
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 ndarray, ndim 1, dtype family int, shape (4,) — shape (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
chdtrc Ran
ran on a synthesized input
AlCorreia/GeFs/gefs/statsutils.py
code served (permissive licence) · get_code("1f4f7b7d0b18ad1a")
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

# The function chdtrc takes two float64 parameters: x and y
# It's a chi-square complementary cumulative distribution function
# from scipy.special, which takes degrees of freedom and chi-square value

# x represents degrees of freedom (must be positive)
# y represents the chi-square value (must be non-negative)

x_val = 5.0  # degrees of freedom
y_val = 2.5  # chi-square value

ARGS = {
    'x': x_val,
    'y': y_val
}
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
get_distance Ran
ran on a synthesized input
AlCorreia/GeFs/gefs/cluster.py
code served (permissive licence) · get_code("6f9a3fa8bb5e3800")
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

# Construct valid arguments for get_distance function
# The function takes three parameters: a, b, and ncat

# Create small numpy arrays for a and b
# These should be 1D arrays of floats
a = np.array([1.0, 2.0, 3.0, 4.0], dtype=np.float64)
b = np.array([1.5, 2.5, 3.5, 4.5], dtype=np.float64)

# Create ncat array indicating number of categories for each feature
# Values > 1 indicate categorical, 1 indicates continuous
ncat = np.array([2, 1, 3, 1], dtype=np.int64)

ARGS = {
    'a': a,
    'b': b,
    'ncat': ncat
}
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
gini Ran
ran on a synthesized input
AlCorreia/GeFs/gefs/split.py
code served (permissive licence) · get_code("22dfc10f1898c675")
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
from numba import njit, int64, float64, optional, prange, deferred_type, types, boolean

@njit(fastmath=True, inline='always')
def gini(left_counts, right_counts, n):
    total_left = 0
    total_right = 0
    gini_left = 0
    gini_right = 0
    for i in range(n):
        gini_left -= left_counts[i] ** 2
        gini_right -= right_counts[i] ** 2
        total_left += left_counts[i]
        total_right += right_counts[i]
    gini_left = total_left - gini_left / total_left
    gini_right = total_right - gini_right / total_right
    return gini_left + gini_right

left_counts = np.array([10, 20, 15, 5], dtype=np.int64)
right_counts = np.array([8, 12, 25, 15], dtype=np.int64)
n = 4

ARGS = {
    'left_counts': left_counts,
    'right_counts': right_counts,
    'n': n
}
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
in1d_scal_nb Ran
ran on a synthesized input
AlCorreia/GeFs/gefs/utils.py
code served (permissive licence) · get_code("73f415a9d6a912c8")
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 1, dtype family float, shape (5,) 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 numba as nb
from numba import njit
import numpy as np

# Create small numpy arrays for testing
matrix = np.array([1, 2, 3, 4, 5], dtype=np.int64)
index_to_remove = np.int64(3)

ARGS = {
    "matrix": matrix,
    "index_to_remove": index_to_remove,
}
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 ndarray, ndim 1, dtype family float, shape (5,) — shape (5,) 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
in1d_vec_nb Ran
ran on a synthesized input
AlCorreia/GeFs/gefs/utils.py
code served (permissive licence) · get_code("50e658133f99d462")
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 1, dtype family float, shape (5,) 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
import numba as nb

# Create a simple 1D numpy array to use as the matrix parameter
# The function checks if each element of matrix is in index_to_remove_set
matrix = np.array([1, 2, 3, 4, 5], dtype=np.int64)

# Create index_to_remove as a numpy array with some values
# The function converts this to a set and checks membership
index_to_remove = np.array([2, 4], dtype=np.int64)

ARGS = {
    "matrix": matrix,
    "index_to_remove": index_to_remove,
}
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 ndarray, ndim 1, dtype family float, shape (5,) — shape (5,) 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
kruskal Ran
ran on a synthesized input
AlCorreia/GeFs/gefs/statsutils.py
code served (permissive licence) · get_code("20c8877af4c4f7ad")
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 numpy as np
import numba as nb

# The kruskal function takes two 1D arrays as input
# Looking at the function body:
# - x and y are concatenated
# - rankdata is called on the concatenated array
# - rank_x and rank_y are extracted from the ranked array
# - sums and lengths are computed from these ranks

# Create two small 1D numpy arrays with some numeric data
x = np.array([1.5, 2.3, 3.1, 4.2], dtype=np.float64)
y = np.array([2.1, 3.5, 4.8, 5.2], dtype=np.float64)

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
numba, numpy — 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
nan_to_num Ran
ran on a synthesized input
AlCorreia/GeFs/gefs/cluster.py
code served (permissive licence) · get_code("e4c4de66b03f27f1")
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 (2, 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
from numba import njit

# Create a numpy array with some NaN values to pass to nan_to_num
data = np.array([[1.0, np.nan, 3.0, 4.0],
                  [np.nan, 2.0, np.nan, 5.0]], dtype=np.float64)

ARGS = {
    'data': data
}
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 ndarray, ndim 2, dtype family float, shape (2, 4) — shape (2, 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
random_split Ran
ran on a synthesized input
AlCorreia/GeFs/gefs/split.py
code served (permissive licence) · get_code("90701f302c52610e")
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 2, 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 numpy as np

np.random.seed(0)

# The function takes a single parameter 'x' which should be an array-like object
# It uses np.random.choice and indexing operations on x
# So x should be a numpy array

x = np.array([1.5, 2.3, 3.7, 4.1, 5.2, 6.8])

ARGS = {
    'x': x
}
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 tuple, ndim 2, 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
resample_strat Ran
ran on a synthesized input
AlCorreia/GeFs/gefs/utils.py
code served (permissive licence) · get_code("14440f4a552431c2")
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

# Create small sample data
np.random.seed(0)

# x: 6 samples, 3 features
x = np.random.randn(6, 3).astype(np.float32)

# y: 6 class labels (3 classes: 0, 1, 2)
y = np.array([0, 0, 1, 1, 2, 2], dtype=np.int64)

# n_classes: number of distinct classes
n_classes = 3

# random_state: for reproducibility
random_state = 42

ARGS = {
    "x": x,
    "y": y,
    "n_classes": n_classes,
    "random_state": random_state,
}
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 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
Leaf Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/nodes.py
code served (permissive licence) · get_code("18680f3973a2a2df")
How it ran

no invocation recorded.

ProdNode Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/nodes.py
code served (permissive licence) · get_code("4d7702c9f7355db1")
How it ran

no invocation recorded.

SumNode Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/nodes.py
code served (permissive licence) · get_code("ef47e92e90a4cb90")
How it ran

no invocation recorded.

add_node Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/learning.py
code served (permissive licence) · get_code("cfe4c67d4efa910e")
How it ran

no invocation recorded.

add_prodnode Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/learning.py
code served (permissive licence) · get_code("cf0b1af1b7139f55")
How it ran

no invocation recorded.

calc_inbag Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/sklearn_utils.py
code served (permissive licence) · get_code("498d492d37875d9a")
How it ran

no invocation recorded.

calc_outofbag Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/sklearn_utils.py
code served (permissive licence) · get_code("41ab67d70fb1d915")
How it ran

no invocation recorded.

cluster Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/cluster.py
code served (permissive licence) · get_code("f0903cda1a853f9f")
How it ran

no invocation recorded.

evaluate Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/trees.py
code served (permissive licence) · get_code("e4dfc5307cff9624")
How it ran

no invocation recorded.

evaluate_instance Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/trees.py
code served (permissive licence) · get_code("fc19aad5d4a58f55")
How it ran

no invocation recorded.

fit Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/learning.py
code served (permissive licence) · get_code("e9e0d52c893cf2bb")
How it ran

no invocation recorded.

signed_join Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/signed.py
code served (permissive licence) · get_code("c6bc9b119c7f3221")
How it ran

no invocation recorded.

signed_prod Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/signed.py
code served (permissive licence) · get_code("65247f439343b294")
How it ran

no invocation recorded.

signed_sum_vec Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/signed.py
code served (permissive licence) · get_code("879ba1caf1941441")
How it ran

no invocation recorded.

vec_chdtrc Not yet run
unverified (harvested, no recorded run)
AlCorreia/GeFs/gefs/statsutils.py
code served (permissive licence) · get_code("50cbc307689e6b0f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 25 samples lifted from this paper's linked repositories, 25 carry a function identity and 0 do not. Of the 25 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 25 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

Decision Trees and Random Forests are among the most widely used machine learning models, and often achieve state-of-the-art performance in tabular, domain-agnostic datasets. Nonetheless, being primarily discriminative models they lack principled methods to manipulate the uncertainty of predictions. In this paper, we exploit Generative Forests (GeFs), a recent class of deep probabilistic models that addresses these issues by extending Random Forests to generative models representing the full joint distribution over the feature space. We demonstrate that GeFs are uncertainty-aware classifiers, capable of measuring the robustness of each prediction as well as detecting out-of-distribution samples.

For agents

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

get_harvested_code_for_paper("2007.05721")
get_code_for_paper("2007.05721")
have("2007.05721")

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

curl https://syntology.ai/api/ran/2007.05721.json

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

[![Syntology run record](https://syntology.ai/api/ran/2007.05721.svg)](https://syntology.ai/paper/2007.05721)

Connect an agent — have() is free.