SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1702.07476 · 2017

Renyi Differential Privacy

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/1702.07476.json · badge: /api/ran/1702.07476.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
donlapark/dirichlet-mechanism community (classifier label)
classifier basis not recorded on the link
8 ran · 0 no recorded run
FunctionStatusWhere it lives
Dlogdelta Ran
ran on a synthesized input
donlapark/dirichlet-mechanism/utils.py
code served (permissive licence) · get_code("cf19378162871d2f")
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 float64, ndim 0, 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
from scipy.special import polygamma, loggamma

# Set random seed for reproducibility
np.random.seed(0)

# Define ARGS dictionary with valid arguments for Dlogdelta
ARGS = {
    "x": 2.5,  # scalar float > 1 (needed for np.log(x-1) and np.log(x))
    "epsilon": 0.5,  # scalar float, privacy parameter
    "r": 1.2,  # scalar float, positive value used in calculations
    "prior": 0.8,  # scalar float, prior value
    "Delta_2sq": 0.1,  # scalar float, squared L2 sensitivity
    "Delta_inf": 0.05,  # scalar float, L-infinity sensitivity
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
Dlogdelta Ran
ran on a synthesized input
donlapark/dirichlet-mechanism/experiments/utils.py
code served (permissive licence) · get_code("933e716d93255b7b")
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 float64, ndim 0, 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
from scipy.special import polygamma, loggamma, digamma

# Set random seed for reproducibility
np.random.seed(0)

# Define ARGS dict with all parameters for Dlogdelta function
ARGS = {
    "x": 2.5,  # scalar float value > 1
    "epsilon": 0.1,  # scalar float value for privacy budget
    "r": 1.5,  # scalar float value for RDP order parameter
    "prior": 0.5,  # scalar float value for prior parameter
    "Delta_2sq": 0.2,  # scalar float value for L2 sensitivity squared
    "Delta_inf": 0.3,  # scalar float value for L-infinity sensitivity
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
compute_vertices Ran
ran on a synthesized input
donlapark/dirichlet-mechanism/experiments/prior_work.py
code served (permissive licence) · get_code("77bd5b2e659c9b6c")
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 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.

from mpmath import beta, gamma, log, re, zeros

torch_imported = False
try:
    import torch
    torch_imported = True
except ImportError:
    pass

# The function compute_vertices takes two parameters: e1 and e2
# Looking at the function body, it performs arithmetic operations on e1 and e2
# These appear to be numeric values (scalars) that are used in comparisons and arithmetic

# e1 and e2 are used in:
# - Comparisons: 1 - 2 * e1 > e2, 1 - e1 - e2 > e1
# - Arithmetic: 1 - 2 * e1, 1 - e1 - e2
# - Assignment to mpmath matrix positions

# We need to provide values that work with mpmath operations
from mpmath import mpf

# Create small numeric values for e1 and e2
e1 = mpf('0.1')
e2 = mpf('0.2')

ARGS = {
    'e1': e1,
    'e2': e2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
mpmath, 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 float — 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
compute_vertices_binary Ran
ran on a synthesized input
donlapark/dirichlet-mechanism/experiments/prior_work.py
code served (permissive licence) · get_code("2bc66f712ad8157a")
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 matrix, 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.

from mpmath import zeros

def compute_vertices_binary(e1, e2):
    if 1 - e1 > e2:
        v = zeros(1, 2)
        v[0] = e1
        v[1] = 1 - e1 - e2
    return v

# Set up arguments
# e1 and e2 are expected to be mpmath numbers
# Based on the function logic: if 1 - e1 > e2, the condition is satisfied
# Let's use e1 = 0.3 and e2 = 0.2
# Then: 1 - 0.3 = 0.7 > 0.2, so condition is True

from mpmath import mpf

e1 = mpf('0.3')
e2 = mpf('0.2')

ARGS = {
    'e1': e1,
    'e2': e2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
mpmath — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type matrix, ndim 1, dtype family float — no output shape is stated: observe_signature.py recorded matrix ndim 0, which disagrees with the fixture run
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
epsilon_func Ran
ran on a synthesized input
donlapark/dirichlet-mechanism/utils.py
code served (permissive licence) · get_code("8029ee1d77fc33b7")
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 float64, ndim 0, 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
from scipy.special import polygamma, loggamma

torch.manual_seed(0)

# Analyze the function:
# - x: appears in operations x**2, so needs numeric type (float)
# - epsilon: scalar compared with subtraction result, needs numeric type
# - prior: used in addition with other values, needs numeric type
# - lambda_: used in multiplication, needs numeric type
# - Delta_2sq: used in multiplication with x**2, needs numeric type
# - Delta_inf: used in multiplication, needs numeric type
# All operations are numeric, so we use floats/arrays

x = np.array(0.5)
epsilon = np.array(0.3)
prior = np.array(2.0)
lambda_ = np.array(1.5)
Delta_2sq = np.array(0.8)
Delta_inf = np.array(0.2)

ARGS = {
    "x": x,
    "epsilon": epsilon,
    "prior": prior,
    "lambda_": lambda_,
    "Delta_2sq": Delta_2sq,
    "Delta_inf": Delta_inf,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, torch — 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 — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
epsilon_func Ran
ran on a synthesized input
donlapark/dirichlet-mechanism/experiments/utils.py
code served (permissive licence) · get_code("fb5043a35a684fcf")
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 float64, ndim 0, 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
from scipy.special import polygamma, loggamma, digamma

torch.manual_seed(0)

# Analyze the function:
# - x: appears in arithmetic operations (lambda_ - 1) * Delta_inf * x and x ** 2
#   so it should be a numeric scalar or array
# - epsilon: subtracted from, should be numeric
# - prior: passed to polygamma as first arg (should be numeric/scalar)
# - lambda_: used in arithmetic, should be numeric scalar
# - Delta_2sq: used in multiplication, should be numeric scalar
# - Delta_inf: used in multiplication, should be numeric scalar
# The function computes: epsilon - 0.5 * x ** 2 * lambda_ * Delta_2sq * polygamma(1, a)
# where a = prior + 3 * (lambda_ - 1) * Delta_inf * x

# Create valid arguments - all scalars for simplicity
x = 0.5
epsilon = 1.2
prior = 2.5
lambda_ = 1.5
Delta_2sq = 0.8
Delta_inf = 0.3

ARGS = {
    "x": x,
    "epsilon": epsilon,
    "prior": prior,
    "lambda_": lambda_,
    "Delta_2sq": Delta_2sq,
    "Delta_inf": Delta_inf,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, torch — 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 — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
logdelta Ran
ran on a synthesized input
donlapark/dirichlet-mechanism/utils.py
code served (permissive licence) · get_code("b08020d271abb428")
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 float64, ndim 0, 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
from scipy.special import polygamma, loggamma

# Set random seed for reproducibility
np.random.seed(0)

# Construct arguments for logdelta function
# The function computes an objective for RDP-to-DP conversion
# It takes scalars for most parameters

x = 1.5  # x should be > 1 (used in x-1 terms)
epsilon = 0.1  # privacy parameter epsilon
r = 0.5  # parameter r (appears in variance calculation)
prior = 0.2  # prior parameter (used in computing 'a')
Delta_2sq = 0.3  # squared L2 sensitivity
Delta_inf = 0.4  # L-infinity sensitivity

ARGS = {
    "x": x,
    "epsilon": epsilon,
    "r": r,
    "prior": prior,
    "Delta_2sq": Delta_2sq,
    "Delta_inf": Delta_inf,
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
logdelta Ran
ran on a synthesized input
donlapark/dirichlet-mechanism/experiments/utils.py
code served (permissive licence) · get_code("525a13e4c8eafc46")
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 float64, ndim 0, 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
from scipy.special import polygamma, loggamma, digamma

# Set random seed for reproducibility
np.random.seed(0)

# Construct arguments for logdelta function
# The function computes an objective function for RDP-to-(epsilon-delta)-DP conversion
# All parameters are numeric scalars or arrays

ARGS = {
    "x": 2.5,  # A scalar value > 1 (used in (x-1) and x*log(x) terms)
    "epsilon": 0.5,  # Privacy budget epsilon (scalar)
    "r": 1.2,  # Noise parameter (scalar)
    "prior": 0.1,  # Prior parameter (scalar)
    "Delta_2sq": 0.3,  # L2 sensitivity squared (scalar)
    "Delta_inf": 0.2,  # L-infinity sensitivity (scalar)
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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

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 5 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 5 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

We propose a natural relaxation of differential privacy based on the Renyi divergence. Closely related notions have appeared in several recent papers that analyzed composition of differentially private mechanisms. We argue that the useful analytical tool can be used as a privacy definition, compactly and accurately representing guarantees on the tails of the privacy loss. We demonstrate that the new definition shares many important properties with the standard definition of differential privacy, while additionally allowing tighter analysis of composite heterogeneous mechanisms.

For agents

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

get_harvested_code_for_paper("1702.07476")
get_code_for_paper("1702.07476")
have("1702.07476")

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

curl https://syntology.ai/api/ran/1702.07476.json

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

[![Syntology run record](https://syntology.ai/api/ran/1702.07476.svg)](https://syntology.ai/paper/1702.07476)

Connect an agent — have() is free.