SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1506.02564 · 2015

Gradient-free Hamiltonian Monte Carlo with Efficient Kernel Exponential Families

arXiv · PDF · Open in the Atlas

Code that ran

21 of 21 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 21 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/1506.02564.json · badge: /api/ran/1506.02564.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
karlnapf/kernel_hmc canonical (classifier label)
classifier basis: census
10 ran · 0 no recorded run
karlnapf/kernel_exp_family community (classifier label)
classifier basis not recorded on the link
11 ran · 0 no recorded run
FunctionStatusWhere it lives
SE Ran
ran on a synthesized input
karlnapf/kernel_exp_family/kernel_exp_family/kernels/develop/kernels.py
code served (permissive licence) · get_code("569bcdd97bd75b7e")
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
import torch

torch.manual_seed(0)

# SE function expects:
# x: array-like, shape (n,) or (n, d)
# y: array-like, shape (n,) or (n, d)
# l: scalar, default=2

# Create small numpy arrays for x and y
x = np.array([1.0, 2.0, 3.0, 4.0])
y = np.array([1.5, 2.5, 3.5, 4.5])
l = 2

ARGS = {
    'x': x,
    'y': y,
    'l': l
}
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 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_C Ran
ran on a synthesized input
karlnapf/kernel_exp_family/kernel_exp_family/estimators/finite/gaussian.py
code served (permissive licence) · get_code("73329303ba1d8c94")
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 (3, 3) 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

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

# Construct arguments for compute_C
# X: 2D array of shape (N, D) where N is number of samples, D is dimension
N = 2
D = 4
X = np.random.randn(N, D)

# omega: 2D array of shape (D, m) where m is the number of random features
m = 3
omega = np.random.randn(D, m)

# u: scalar or 1D array of length m (bias term)
u = np.random.randn(m)

ARGS = {
    "X": X,
    "omega": omega,
    "u": u,
}
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 ndarray, ndim 2, dtype family float, shape (3, 3) — shape (3, 3) 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
compute_C Ran
ran on a synthesized input
karlnapf/kernel_exp_family/kernel_exp_family/estimators/lite/gaussian.py
code served (permissive licence) · get_code("cc2693e2c6f2983a")
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, 2) 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

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

# Create small test data
# X: shape (NX, D) - NX samples, D features
# Y: shape (NY, D) - NY samples, D features
# K: shape (NX, NY) - kernel matrix

NX = 2  # number of X samples
NY = 3  # number of Y samples
D = 4   # feature dimension

X = np.random.randn(NX, D)
Y = np.random.randn(NY, D)
K = np.random.randn(NX, NY)
sigma = 0.5

ARGS = {
    "X": X,
    "Y": Y,
    "K": K,
    "sigma": sigma
}
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 ndarray, ndim 2, dtype family float, shape (2, 2) — shape (2, 2) 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
compute_C_sym Ran
ran on a synthesized input
karlnapf/kernel_exp_family/kernel_exp_family/estimators/lite/develop/gaussian.py
code served (permissive licence) · get_code("3dd923a3df120950")
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, 2) 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

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

# Create small test arrays
# Z: N x D matrix where N=2 (number of samples), D=4 (feature dimension)
Z = np.random.randn(2, 4)

# K: N x N symmetric matrix (kernel matrix)
K = np.random.randn(2, 2)
K = (K + K.T) / 2  # Make it symmetric

# sigma: scalar parameter
sigma = 0.5

ARGS = {
    "Z": Z,
    "K": K,
    "sigma": sigma
}
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 ndarray, ndim 2, dtype family float, shape (2, 2) — shape (2, 2) 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
compute_b Ran
ran on a synthesized input
karlnapf/kernel_exp_family/kernel_exp_family/estimators/lite/gaussian.py
code served (permissive licence) · get_code("f0de675df2ee5b85")
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 (2,) 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

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

# Create small sample data
# X: (NX, D) where NX=2, D=4
X = np.random.randn(2, 4).astype(np.float64)

# Y: (NY, D) where NY=3, D=4
Y = np.random.randn(3, 4).astype(np.float64)

# K_XY: (NX, NY) kernel matrix
K_XY = np.random.randn(2, 3).astype(np.float64)

# sigma: scalar parameter
sigma = 1.5

ARGS = {
    "X": X,
    "Y": Y,
    "K_XY": K_XY,
    "sigma": sigma,
}
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 ndarray, ndim 1, dtype family float, shape (2,) — shape (2,) 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
compute_b_sym Ran
ran on a synthesized input
karlnapf/kernel_exp_family/kernel_exp_family/estimators/lite/develop/gaussian.py
code served (permissive licence) · get_code("b4442e480a0f9c3d")
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 (2,) 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 small numpy arrays with distinct dimensions
N = 2  # number of samples
D = 4  # number of dimensions

# Z: shape (N, D) - data matrix
Z = np.random.randn(N, D)

# K: shape (N, N) - kernel/covariance matrix
K = np.random.randn(N, N)

# sigma: scalar - variance parameter
sigma = 1.5

ARGS = {
    "Z": Z,
    "K": K,
    "sigma": sigma,
}
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 ndarray, ndim 1, dtype family float, shape (2,) — shape (2,) 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
compute_hamiltonian Ran
ran on a synthesized input
karlnapf/kernel_hmc/kernel_hmc/hamiltonian/hamiltonian.py
code served (permissive licence) · get_code("29e20443e95e1224")
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 (2,) 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

# Define small callable functions for logq and logp
def logq(Q):
    """Log probability for Q"""
    if isinstance(Q, np.ndarray):
        return np.sum(Q ** 2)
    return 0.0

def logp(P):
    """Log probability for P"""
    if isinstance(P, np.ndarray):
        return np.sum(P ** 2)
    return 0.0

# Create small sample data
Qs = [np.array([1.0, 2.0, 3.0, 4.0]), np.array([0.5, 1.5, 2.5, 3.5])]
Ps = [np.array([0.1, 0.2, 0.3, 0.4]), np.array([0.2, 0.3, 0.4, 0.5])]

ARGS = {
    "Qs": Qs,
    "Ps": Ps,
    "logq": logq,
    "logp": logp,
}
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 ndarray, ndim 1, dtype family float, shape (2,) — shape (2,) 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
compute_log_accept_pr Ran
ran on a synthesized input
karlnapf/kernel_hmc/kernel_hmc/hamiltonian/hamiltonian.py
code served (permissive licence) · get_code("277c7ba269a4d872")
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 (2,) 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

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

# Define logq: log probability of q
# Takes a single sample or batch and returns log probability
def logq(Q):
    # Q can be shape (dim,) or (batch, dim)
    if Q.ndim == 1:
        return np.sum(Q ** 2)  # simple quadratic log prob
    else:
        return np.sum(Q ** 2, axis=1)  # batch-wise sum

# Define logp: log probability of p
# Takes a single sample or batch and returns log probability
def logp(P):
    # P can be shape (dim,) or (batch, dim)
    if P.ndim == 1:
        return np.sum(P ** 2)  # simple quadratic log prob
    else:
        return np.sum(P ** 2, axis=1)  # batch-wise sum

# Create initial state q0 and p0 (dimension 4)
q0 = np.random.randn(4).astype(np.float32)
p0 = np.random.randn(4).astype(np.float32)

# Create batch of Qs and Ps (batch size 2, dimension 4)
Qs = np.random.randn(2, 4).astype(np.float32)
Ps = np.random.randn(2, 4).astype(np.float32)

ARGS = {
    'q0': q0,
    'p0': p0,
    'Qs': Qs,
    'Ps': Ps,
    'logq': logq,
    'logp': logp,
}
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 ndarray, ndim 1, dtype family float, shape (2,) — shape (2,) 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
compute_log_accept_pr_single Ran
ran on a synthesized input
karlnapf/kernel_hmc/kernel_hmc/hamiltonian/hamiltonian.py
code served (permissive licence) · get_code("e9d1b01844c6fdbe")
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 float32, 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

# Define simple log probability functions
def logq(q):
    """Log probability of q (data/position space)"""
    # Assume q is a vector, return a scalar log prob
    return -0.5 * np.sum(q**2)

def logp(p):
    """Log probability of p (momentum space)"""
    # Assume p is a vector, return a scalar log prob
    return -0.5 * np.sum(p**2)

# Create small test arrays with dimension 4
np.random.seed(0)
q0 = np.random.randn(4).astype(np.float32)
p0 = np.random.randn(4).astype(np.float32)
q = np.random.randn(4).astype(np.float32)
p = np.random.randn(4).astype(np.float32)

ARGS = {
    "q0": q0,
    "p0": p0,
    "q": q,
    "p": p,
    "logq": logq,
    "logp": logp,
}
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 float32, 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
compute_lower_right_submatrix_loop Ran
ran on a synthesized input
karlnapf/kernel_exp_family/kernel_exp_family/estimators/full/develop/gaussian.py
code served (permissive licence) · get_code("4cc823ebc7943656")
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 (8, 8) 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

def kernel_dx_dy(x, y):
    """
    Stand-in kernel function that takes two column vectors and returns a d x d matrix.
    x: shape (d, 1)
    y: shape (d, 1)
    Returns: (d, d) matrix
    """
    d = x.shape[0]
    # Simple kernel: outer product scaled by a constant
    result = np.dot(x, y.T) + np.eye(d)
    return result

# Create small data: n=2 samples, d=4 dimensions
n = 2
d = 4
data = np.random.randn(n, d).astype(np.float64)

# Lambda parameter for regularization
lmbda = 0.1

ARGS = {
    "kernel_dx_dy": kernel_dx_dy,
    "data": data,
    "lmbda": lmbda,
}
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 ndarray, ndim 2, dtype family float, shape (8, 8) — shape (8, 8) 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
gamma_median_heuristic Ran
ran on a synthesized input
karlnapf/kernel_hmc/kernel_hmc/proposals/metropolis.py
code served (permissive licence) · get_code("e07fd07ecd9bf03e")
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.spatial.distance import cdist, squareform, pdist

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

# Create a small sample dataset Z
# Z should be a 2D array where each row is a data point
# Using shape (10, 4) - 10 samples, 4 dimensions
Z = np.random.randn(10, 4).astype(np.float64)

# Define ARGS dict with all parameters
ARGS = {
    "Z": Z,
    "num_subsample": 5
}
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
gaussian_kernel Ran
ran on a synthesized input
karlnapf/kernel_exp_family/kernel_exp_family/kernels/kernels.py
code served (permissive licence) · get_code("1a4110008b634268")
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, 3) 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 scipy.spatial.distance

# Create small numpy arrays for testing
X = np.array([[1.0, 2.0, 3.0, 4.0],
              [5.0, 6.0, 7.0, 8.0]], dtype=np.float64)

Y = np.array([[0.5, 1.5, 2.5, 3.5],
              [4.5, 5.5, 6.5, 7.5],
              [9.0, 10.0, 11.0, 12.0]], dtype=np.float64)

sigma = 1.0

ARGS = {
    'X': X,
    'Y': Y,
    'sigma': sigma
}
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 ndarray, ndim 2, dtype family float, shape (2, 3) — shape (2, 3) 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
gaussian_kernel_dx_component Ran
ran on a synthesized input
karlnapf/kernel_exp_family/kernel_exp_family/kernels/kernels.py
code served (permissive licence) · get_code("8bd2226b9a61cb5b")
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
import scipy.spatial.distance

# Create small 1D vectors for x and y
x = np.array([1.0, 2.0, 3.0, 4.0])
y = np.array([1.5, 2.5, 3.5, 4.5])

# ell parameter - index into the vectors
ell = 2

# sigma parameter - bandwidth for gaussian kernel
sigma = 1.0

ARGS = {
    "x": x,
    "y": y,
    "ell": ell,
    "sigma": sigma
}
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
gaussian_kernel_grad Ran
ran on a synthesized input
karlnapf/kernel_exp_family/kernel_exp_family/kernels/kernels.py
code served (permissive licence) · get_code("e1b95660ade3b969")
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
import scipy.spatial.distance

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

# Create a small 1D input vector (single point)
x = np.random.randn(4)

# Create a small 2D array of points (batch of 2 points, each with 4 dimensions)
Y = np.random.randn(2, 4)

# Set sigma parameter
sigma = 1.0

# Construct the arguments dictionary
ARGS = {
    'x': x,
    'Y': Y,
    'sigma': sigma
}
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 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
leapfrog Ran
ran on a synthesized input
karlnapf/kernel_hmc/kernel_hmc/hamiltonian/leapfrog.py
code served (permissive licence) · get_code("6c764cff71d22a7d")
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 3, 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

# Define gradient functions for dlogq and dlogp
def dlogq_func(q):
    """Gradient of log q - simple quadratic potential"""
    return 2 * q

def dlogp_func(p):
    """Gradient of log p - simple quadratic kinetic energy"""
    return p

# Initial position and momentum
q_init = np.array([1.0, 2.0, 3.0, 4.0])
p_init = np.array([0.5, 0.5, 0.5, 0.5])

ARGS = {
    "q": q_init,
    "dlogq": dlogq_func,
    "p": p_init,
    "dlogp": dlogp_func,
    "step_size": 0.3,
    "num_steps": 1
}
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, ndim 3, 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
leapfrog_no_storing Ran
ran on a synthesized input
karlnapf/kernel_hmc/kernel_hmc/hamiltonian/leapfrog.py
code served (permissive licence) · get_code("6fb85dfc6b7b817e")
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

def dlogq_func(q):
    """Gradient of log probability w.r.t. q"""
    return np.array([0.1, 0.2, 0.3, 0.4])

def dlogp_func(p):
    """Gradient of log probability w.r.t. p"""
    return np.array([0.05, 0.15, 0.25, 0.35])

q_init = np.array([1.0, 2.0, 3.0, 4.0])
p_init = np.array([0.5, 1.0, 1.5, 2.0])

ARGS = {
    "q": q_init,
    "dlogq": dlogq_func,
    "p": p_init,
    "dlogp": dlogp_func,
    "step_size": 0.3,
    "num_steps": 1
}
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, 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
log_gaussian_pdf Ran
ran on a synthesized input
karlnapf/kernel_hmc/kernel_hmc/densities/gaussian.py
code served (permissive licence) · get_code("fae9cdd794af0134")
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.linalg import solve_triangular

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

# Create a small covariance matrix (4x4)
dim = 4
Sigma = np.random.randn(dim, dim)
Sigma = Sigma @ Sigma.T  # Make it positive definite and symmetric

# Compute Cholesky decomposition
L = np.linalg.cholesky(Sigma)

# Create mean vector
mu = np.random.randn(dim)

# Create data point x
x = np.random.randn(dim)

# Define ARGS dict with all parameters
ARGS = {
    "x": x,
    "mu": mu,
    "Sigma": Sigma,
    "is_cholesky": False,
    "compute_grad": False,
}
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
rank_update_mean_covariance_cholesky_lmbda_naive Ran
ran on a synthesized input
karlnapf/kernel_hmc/kernel_hmc/proposals/metropolis.py
code served (permissive licence) · get_code("de55b9d7d9628e5e")
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

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

# Construct arguments for rank_update_mean_covariance_cholesky_lmbda_naive
D = 4

# u: 1D array of dimension D
u = np.random.randn(D)

# lmbda: interpolation parameter between 0 and 1
lmbda = 0.3

# mean: 1D array of dimension D (optional, can be None)
mean = np.random.randn(D)

# cov_L: lower Cholesky factor (D x D matrix)
# Create a valid Cholesky factor by constructing a positive definite matrix
# and taking its Cholesky decomposition
temp_cov = np.eye(D) + np.random.randn(D, D) * 0.1
temp_cov = temp_cov @ temp_cov.T  # Make it positive definite
cov_L = np.linalg.cholesky(temp_cov)

# nu2: scaling parameter for outer product
nu2 = 1.0

# gamma2: optional isotropic term (can be None or a positive float)
gamma2 = 0.05

ARGS = {
    "u": u,
    "lmbda": lmbda,
    "mean": mean,
    "cov_L": cov_L,
    "nu2": nu2,
    "gamma2": gamma2,
}
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
rff_feature_map_grad_loop Ran
ran on a synthesized input
karlnapf/kernel_exp_family/kernel_exp_family/kernels/develop/kernels.py
code served (permissive licence) · get_code("b18f89dae51889f8")
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 3, dtype family float, shape (4, 2, 3) 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

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

# Parameters for rff_feature_map_grad_loop:
# X: input data array of shape (N, D) where N is number of samples, D is dimension
# omega: weight matrix of shape (D, m) where m is number of features
# u: bias term, can be scalar or array of length m

N = 2  # number of samples
D = 4  # input dimension
m = 3  # number of RFF features

# X: (N, D) array
X = np.random.randn(N, D)

# omega: (D, m) array - the function does np.dot(X, omega) so omega must be (D, m)
omega = np.random.randn(D, m)

# u: can be scalar or array of length m
u = np.random.randn(m)

ARGS = {
    "X": X,
    "omega": omega,
    "u": u,
}
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 ndarray, ndim 3, dtype family float, shape (4, 2, 3) — shape (4, 2, 3) 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
sample_gaussian Ran
ran on a synthesized input
karlnapf/kernel_hmc/kernel_hmc/densities/gaussian.py
code served (permissive licence) · get_code("ff81502c713bcd2f")
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 (5, 2) 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

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

# Define arguments for sample_gaussian
N = 5  # Number of samples to draw
mu = np.array([0.0, 0.0])  # Mean vector of dimension 2
Sigma = np.eye(2)  # Covariance matrix (2x2 identity)
is_cholesky = False  # Flag indicating Sigma is not already Cholesky decomposed

ARGS = {
    "N": N,
    "mu": mu,
    "Sigma": Sigma,
    "is_cholesky": is_cholesky
}
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 ndarray, ndim 2, dtype family float, shape (5, 2) — shape (5, 2) 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
standard_sqrt_schedule Ran
ran on a synthesized input
karlnapf/kernel_hmc/kernel_hmc/proposals/base.py
code served (permissive licence) · get_code("92bf96ecad7e3e6f")
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

def standard_sqrt_schedule(t):
    return 1.0 / np.sqrt(t + 1)

# The function takes a single parameter 't' which is used in a mathematical operation
# t can be an integer, float, or numpy array
# Based on the function body, it performs: 1.0 / np.sqrt(t + 1)

ARGS = {
    't': 5
}
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 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 21 samples lifted from this paper's linked repositories, 21 carry a function identity and 0 do not. Of the 21 distinct functions with an identity, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 2 later) and 19 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
leapfrog 1 other paper
1 later
1711.05363 (2017-11)
Kernel Conditional Exponential Family
standard_sqrt_schedule 1 other paper
1 later
1711.05363 (2017-11)
Kernel Conditional Exponential Family

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 Kernel Hamiltonian Monte Carlo (KMC), a gradient-free adaptive MCMC algorithm based on Hamiltonian Monte Carlo (HMC). On target densities where classical HMC is not an option due to intractable gradients, KMC adaptively learns the target's gradient structure by fitting an exponential family model in a Reproducing Kernel Hilbert Space. Computational costs are reduced by two novel efficient approximations to this gradient. While being asymptotically exact, KMC mimics HMC in terms of sampling efficiency, and offers substantial mixing improvements over state-of-the-art gradient free samplers. We support our claims with experimental studies on both toy and real-world applications, including Approximate Bayesian Computation and exact-approximate MCMC.

For agents

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

get_harvested_code_for_paper("1506.02564")
get_code_for_paper("1506.02564")
have("1506.02564")

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

curl https://syntology.ai/api/ran/1506.02564.json

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

[![Syntology run record](https://syntology.ai/api/ran/1506.02564.svg)](https://syntology.ai/paper/1506.02564)

Connect an agent — have() is free.