SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1806.04326 · 2018

Differentiable Compositional Kernel Learning for Gaussian Processes

arXiv · PDF · Open in the Atlas

Code that ran

8 of 15 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 7 have no contract verdict; 1 failure of our instrument; 7 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/1806.04326.json · badge: /api/ran/1806.04326.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
thjashin/spectral-stein-grad canonical (classifier label)
classifier basis: census
7 ran · 5 no recorded run
ssydasheng/Neural-Kernel-Network canonical (classifier label)
classifier basis: classifier_2025
1 ran · 2 no recorded run
FunctionStatusWhere it lives
average_gradients Ran
ran on a synthesized input
thjashin/spectral-stein-grad/utils/multi_gpu.py
code served (permissive licence) · get_code("19a2e9b1682d85e8")
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. 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 tensorflow as tf
import numpy as np

# Create sample tower gradients
# Structure: List of lists of (gradient, variable) tuples
# Outer list: over individual variables/gradients
# Inner list: over towers

# Create a simple variable (tensor) to pair with gradients
var1 = tf.Variable(tf.random.normal([4, 3]))
var2 = tf.Variable(tf.random.normal([2, 5]))

# Create gradients for 2 towers and 2 variables
# Tower 1 gradients
tower1_grads = [
    (tf.random.normal([4, 3]), var1),
    (tf.random.normal([2, 5]), var2),
]

# Tower 2 gradients
tower2_grads = [
    (tf.random.normal([4, 3]), var1),
    (tf.random.normal([2, 5]), var2),
]

# Tower 3 gradients
tower3_grads = [
    (tf.random.normal([4, 3]), var1),
    (tf.random.normal([2, 5]), var2),
]

# Structure: list of lists of (gradient, variable) tuples
tower_grads = [tower1_grads, tower2_grads, tower3_grads]

ARGS = {
    "tower_grads": tower_grads
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, six, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list — a list 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
average_losses Ran
ran on a synthesized input
thjashin/spectral-stein-grad/utils/multi_gpu.py
code served (permissive licence) · get_code("b68141eeaf8d69a1")
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 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 tensorflow as tf
import numpy as np

# Create sample tower losses
# The function expects a list of lists, where outer list is towers,
# inner list is losses/quantities for each tower
tower_1_losses = [
    tf.constant(1.5, dtype=tf.float32),
    tf.constant(2.0, dtype=tf.float32),
    tf.constant(0.5, dtype=tf.float32)
]

tower_2_losses = [
    tf.constant(1.3, dtype=tf.float32),
    tf.constant(2.2, dtype=tf.float32),
    tf.constant(0.6, dtype=tf.float32)
]

tower_3_losses = [
    tf.constant(1.4, dtype=tf.float32),
    tf.constant(1.9, dtype=tf.float32),
    tf.constant(0.4, dtype=tf.float32)
]

tower_losses = [tower_1_losses, tower_2_losses, tower_3_losses]

ARGS = {
    "tower_losses": tower_losses
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, six, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, 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
calculate_frechet_distance Ran
ran on a synthesized input
thjashin/spectral-stein-grad/utils/fid.py
code served (permissive licence) · get_code("48918e785b59da5c")
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 as sp

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

# Create small sample means (feature vectors)
# Using dimension 64 as typical for inception pool_3 features
mu1 = np.random.randn(64).astype(np.float64)
mu2 = np.random.randn(64).astype(np.float64)

# Create small positive semi-definite covariance matrices
# Generate random matrices and make them symmetric positive semi-definite
A1 = np.random.randn(64, 64).astype(np.float64)
sigma1 = np.dot(A1, A1.T) + np.eye(64) * 0.1  # Add small constant for numerical stability

A2 = np.random.randn(64, 64).astype(np.float64)
sigma2 = np.dot(A2, A2.T) + np.eye(64) * 0.1  # Add small constant for numerical stability

ARGS = {
    'mu1': mu1,
    'sigma1': sigma1,
    'mu2': mu2,
    'sigma2': sigma2,
}
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
entropy_surrogate Ran
ran on a synthesized input
thjashin/spectral-stein-grad/estimator/entropy.py
code served (permissive licence) · get_code("c88441fa4ae915dc")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned EagerTensor, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np


class MockEstimator:
    """Minimal estimator that implements compute_gradients method."""
    
    def compute_gradients(self, samples):
        """
        Compute gradients (log q derivatives).
        Returns a tensor of same shape as samples.
        """
        # Return a tensor with the same shape as samples
        # representing -dlog_q (gradient of log probability)
        batch_size = tf.shape(samples)[0]
        feature_dim = tf.shape(samples)[1]
        return tf.random.normal([batch_size, feature_dim])


# Create sample data: batch_size=2, feature_dim=4
samples = tf.constant([
    [0.1, 0.2, 0.3, 0.4],
    [0.5, 0.6, 0.7, 0.8]
], dtype=tf.float32)

# Create estimator instance
estimator = MockEstimator()

ARGS = {
    "estimator": estimator,
    "samples": samples
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_mnist_realval Ran
ran on a synthesized input
thjashin/spectral-stein-grad/utils/dataset.py
code served (permissive licence) · get_code("72d5c82262942d8c")
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 os
import gzip
import numpy as np
import tempfile
import pickle
import six

# Create a temporary directory for the mock dataset
temp_dir = tempfile.mkdtemp()
mock_path = os.path.join(temp_dir, "mnist.pkl.gz")

# Create a minimal mock MNIST dataset
# Each sample is 784-dimensional (28x28), labels are 0-9
np.random.seed(0)
n_train, n_valid, n_test = 10, 5, 3
x_train = np.random.rand(n_train, 784).astype('float32')
t_train = np.array([i % 10 for i in range(n_train)], dtype=np.int64)
x_valid = np.random.rand(n_valid, 784).astype('float32')
t_valid = np.array([i % 10 for i in range(n_valid)], dtype=np.int64)
x_test = np.random.rand(n_test, 784).astype('float32')
t_test = np.array([i % 10 for i in range(n_test)], dtype=np.int64)

train_set = (x_train, t_train)
valid_set = (x_valid, t_valid)
test_set = (x_test, t_test)

# Write the mock dataset to a gzip pickle file
with gzip.open(mock_path, 'wb') as f:
    pickle.dump((train_set, valid_set, test_set), f)

ARGS = {
    'path': mock_path,
    'one_hot': True,
    'dequantify': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, six — 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
organize_symbol Ran
ran; our contract draft was wrong, not the code
ssydasheng/Neural-Kernel-Network/exp/time-series.py
code served (permissive licence) · get_code("bdf54a85518a5fec")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) 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 sympy as sp

# Create symbolic variables
x, y, z = sp.symbols('x y z')

# Create a symbolic expression that will expand and have coefficients
# For example: (x + 2)^2 = x^2 + 4*x + 4
sym = (x + 2)**2 + 3*y - 0.00005*z + 1.5

ARGS = {
    'sym': sym
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
sympy — 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
standardize Ran
ran on a synthesized input
thjashin/spectral-stein-grad/utils/dataset.py
code served (permissive licence) · get_code("6d66aeae14f49d89")
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 training data: 2 samples, 4 features
data_train = np.array([
    [1.0, 2.0, 3.0, 4.0],
    [5.0, 6.0, 7.0, 8.0]
], dtype=np.float32)

# Create small test data: 2 samples, 4 features
data_test = np.array([
    [2.0, 3.0, 4.0, 5.0],
    [6.0, 7.0, 8.0, 9.0]
], dtype=np.float32)

ARGS = {
    "data_train": data_train,
    "data_test": data_test
}
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
to_one_hot Ran
ran on a synthesized input
thjashin/spectral-stein-grad/utils/dataset.py
code served (permissive licence) · get_code("c28e0523d961eb32")
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 (6, 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

# Create a 1-D numpy array of integers representing class indices
x = np.array([0, 2, 1, 3, 0, 1], dtype=int)

# Define the depth (number of classes)
depth = 4

ARGS = {
    "x": x,
    "depth": depth
}
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 (6, 4) — shape (6, 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
NKNInfo Not yet run
unverified (harvested, no recorded run)
ssydasheng/Neural-Kernel-Network/exp/regression.py
code served (permissive licence) · get_code("6a9551ffe4100ac1")
How it ran

no invocation recorded.

NKNInfo Not yet run
unverified (harvested, no recorded run)
ssydasheng/Neural-Kernel-Network/exp/bayes-opt.py
code served (permissive licence) · get_code("f5cbda4fc6a74c0e")
How it ran

no invocation recorded.

calculate_activation_statistics Not yet run
unverified (harvested, no recorded run)
thjashin/spectral-stein-grad/utils/fid.py
code served (permissive licence) · get_code("6917ffcb1214d74f")
How it ran

no invocation recorded.

create_session Not yet run
unverified (harvested, no recorded run)
thjashin/spectral-stein-grad/utils/utils.py
code served (permissive licence) · get_code("cc6eb1f37d644971")
How it ran

no invocation recorded.

entropy_gradients Not yet run
unverified (harvested, no recorded run)
thjashin/spectral-stein-grad/estimator/entropy.py
code served (permissive licence) · get_code("bfdb3dc0e3c8c2f5")
How it ran

no invocation recorded.

get_activations Not yet run
unverified (harvested, no recorded run)
thjashin/spectral-stein-grad/utils/fid.py
code served (permissive licence) · get_code("e69fcd6206886931")
How it ran

no invocation recorded.

minimize_entropy Not yet run
unverified (harvested, no recorded run)
thjashin/spectral-stein-grad/estimator/entropy.py
code served (permissive licence) · get_code("fe17c7de386fd13d")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 2 later) and 13 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
calculate_activation_statistics 12 other papers
4 dated earlier, 8 later
1406.2661 (2014-06)
Generative Adversarial Networks
1703.10593 (2017-03)
Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
1710.10196 (2017-10)
Progressive Growing of GANs for Improved Quality, Stability, and Variation
and 9 more not named here
get_activations 9 other papers
3 dated earlier, 6 later
1703.10593 (2017-03)
Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
1710.10196 (2017-10)
Progressive Growing of GANs for Improved Quality, Stability, and Variation
1802.10551 (2018-02)
A Variational Inequality Perspective on Generative Adversarial Networks
and 6 more not named here

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

The generalization properties of Gaussian processes depend heavily on the choice of kernel, and this choice remains a dark art. We present the Neural Kernel Network (NKN), a flexible family of kernels represented by a neural network. The NKN architecture is based on the composition rules for kernels, so that each unit of the network corresponds to a valid kernel. It can compactly approximate compositional kernel structures such as those used by the Automatic Statistician (Lloyd et al., 2014), but because the architecture is differentiable, it is end-to-end trainable with gradient-based optimization. We show that the NKN is universal for the class of stationary kernels. Empirically we demonstrate pattern discovery and extrapolation abilities of NKN on several tasks that depend crucially on identifying the underlying structure, including time series and texture extrapolation, as well as Bayesian optimization.

For agents

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

get_harvested_code_for_paper("1806.04326")
get_code_for_paper("1806.04326")
have("1806.04326")

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

curl https://syntology.ai/api/ran/1806.04326.json

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

[![Syntology run record](https://syntology.ai/api/ran/1806.04326.svg)](https://syntology.ai/paper/1806.04326)

Connect an agent — have() is free.