SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1807.06653 · 2018

Invariant Information Clustering for Unsupervised Image Classification and Segmentation

arXiv · PDF · Open in the Atlas

Code that ran

11 of 18 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 9 have no contract verdict; 2 failures 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/1807.06653.json · badge: /api/ran/1807.06653.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
xu-ji/IIC canonical (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
astirn/IIC community (classifier label)
classifier basis not recorded on the link
3 ran · 4 no recorded run
nathanin/IIC community (classifier label)
classifier basis not recorded on the link
3 ran · 2 no recorded run
willi-menapace/acids-clustering-domain-shift community (classifier label)
classifier basis not recorded on the link
3 ran · 1 no recorded run
FunctionStatusWhere it lives
IID_alpha_plus_loss Ran
ran on a synthesized input
willi-menapace/acids-clustering-domain-shift/codebase/utils/cluster/IID_losses.py
code served (permissive licence) · get_code("d9b2528f3dc09ca2")
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 torch
import sys

torch.manual_seed(0)

# Create small tensors for plain_output and transformed_output
# batch_size=2, k=4 (number of classes/clusters)
batch_size = 2
k = 4

plain_output = torch.randn(batch_size, k)
plain_output = torch.nn.functional.softmax(plain_output, dim=1)  # Normalize to be valid probabilities

transformed_output = torch.randn(batch_size, k)
transformed_output = torch.nn.functional.softmax(transformed_output, dim=1)  # Normalize to be valid probabilities

# Create old_joint_probabilities as a k x k matrix
old_joint_probabilities = torch.randn(k, k).abs()
old_joint_probabilities = old_joint_probabilities / old_joint_probabilities.sum()  # Normalize

# Set alpha parameter
alpha = 0.5

# Set lamb parameter
lamb = 1.0

# EPS is optional and has a default value
EPS = sys.float_info.epsilon

ARGS = {
    "plain_output": plain_output,
    "transformed_output": transformed_output,
    "old_joint_probabilities": old_joint_probabilities,
    "alpha": alpha,
    "lamb": lamb,
    "EPS": EPS
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — 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
IID_loss Ran
ran; our contract draft was wrong, not the code
xu-ji/IIC/code/utils/cluster/IID_losses.py
code served (permissive licence) · get_code("b2399f01716e7765")
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, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import sys

torch.manual_seed(0)

# The function expects x_out and x_tf_out to be 2D tensors of shape (bn, k)
# where bn is batch size and k is number of clusters/classes
# They should represent probability distributions (softmax outputs)

bn = 2  # batch size
k = 4   # number of clusters

# Create softmax outputs (probability distributions)
x_out = torch.softmax(torch.randn(bn, k), dim=1)
x_tf_out = torch.softmax(torch.randn(bn, k), dim=1)

# lamb is a scalar weight parameter
lamb = 1.0

# EPS is epsilon for numerical stability
EPS = sys.float_info.epsilon

ARGS = {
    'x_out': x_out,
    'x_tf_out': x_tf_out,
    'lamb': lamb,
    'EPS': EPS
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
IID_loss Ran
ran on a synthesized input
nathanin/IIC/tensorflow/IID_losses.py
code served (permissive licence) · get_code("eafc9f12c5ea5696")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import sys

torch.manual_seed(0)

# Create small tensors for x_out and x_tf_out
# Function expects: x_out.size() = (bn, k) and x_tf_out.size() = (bn, k)
# Let bn=2 (batch size), k=4 (number of clusters/dimensions)
bn = 2
k = 4

x_out = torch.rand(bn, k)
x_tf_out = torch.rand(bn, k)

# Optional parameters
lamb = 1.0
EPS = sys.float_info.epsilon

ARGS = {
    'x_out': x_out,
    'x_tf_out': x_tf_out,
    'lamb': lamb,
    'EPS': EPS
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
IID_loss Ran
ran on a synthesized input
nathanin/IIC/tensorflow/IID_losses_tf.py
code served (permissive licence) · get_code("aeaf2fa6a05b6904")
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 sys

# Set seed for reproducibility
tf.random.set_seed(0)

# Create small test tensors
# x_out: shape (batch_size, feature_dim) - batch of 2, features of 4
batch_size = 2
feature_dim = 4

x_out = tf.random.normal((batch_size, feature_dim), dtype=tf.float32)
x_tf_out = tf.random.normal((batch_size, feature_dim), dtype=tf.float32)

# lamb and EPS parameters
lamb = 1.0
EPS = sys.float_info.epsilon

ARGS = {
    "x_out": x_out,
    "x_tf_out": x_tf_out,
    "lamb": lamb,
    "EPS": EPS
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
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
IID_plus_loss Ran
ran on a synthesized input
willi-menapace/acids-clustering-domain-shift/codebase/utils/cluster/IID_losses.py
code served (permissive licence) · get_code("1217b85cf4046007")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import sys

torch.manual_seed(0)

# Parameters for IID_plus_loss
batch_size = 2
num_classes = 4

# plain_output: shape (batch_size, num_classes), should be probabilities
plain_output = torch.rand(batch_size, num_classes)
plain_output = plain_output / plain_output.sum(dim=1, keepdim=True)

# transformed_output: shape (batch_size, num_classes), should be probabilities
transformed_output = torch.rand(batch_size, num_classes)
transformed_output = transformed_output / transformed_output.sum(dim=1, keepdim=True)

# joint_probabilities_estimations: list of estimated probability matrices
# Each matrix should be shape (num_classes, num_classes)
joint_prob_matrix_1 = torch.rand(num_classes, num_classes)
joint_prob_matrix_1 = joint_prob_matrix_1 / joint_prob_matrix_1.sum()

joint_prob_matrix_2 = torch.rand(num_classes, num_classes)
joint_prob_matrix_2 = joint_prob_matrix_2 / joint_prob_matrix_2.sum()

joint_probabilities_estimations = [joint_prob_matrix_1, joint_prob_matrix_2]

# lamb parameter
lamb = 1.0

# EPS parameter
EPS = sys.float_info.epsilon

ARGS = {
    "plain_output": plain_output,
    "transformed_output": transformed_output,
    "joint_probabilities_estimations": joint_probabilities_estimations,
    "lamb": lamb,
    "EPS": EPS
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
compute_cluster_confusion_matrix Ran
ran on a synthesized input
willi-menapace/acids-clustering-domain-shift/codebase/utils/cluster/cluster_eval.py
code served (permissive licence) · get_code("66416b36a224f011")
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

# Create small example data for clustering evaluation
# We'll have 10 samples with 3 predicted clusters and 3 ground truth classes
np.random.seed(0)

# Cluster assignments: which cluster each sample belongs to (values 0 to cluster_k-1)
clusters = np.array([0, 0, 1, 1, 2, 2, 0, 1, 2, 0], dtype=int)

# Ground truth cluster assignments (values 0 to ground_truth_k-1)
ground_truth_clusters = np.array([0, 0, 1, 1, 2, 2, 0, 1, 2, 0], dtype=int)

# Number of clusters predicted
cluster_k = 3

# Number of ground truth classes
ground_truth_k = 3

ARGS = {
    "clusters": clusters,
    "ground_truth_clusters": ground_truth_clusters,
    "cluster_k": cluster_k,
    "ground_truth_k": ground_truth_k,
}
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_joint Ran
ran; our contract draft was wrong, not the code
xu-ji/IIC/code/utils/cluster/IID_losses.py
code served (permissive licence) · get_code("76deff4b3537c768")
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 Tensor, ndim 2, dtype family float, shape (4, 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 torch

torch.manual_seed(0)

bn = 2
k = 4

# Create two tensors with shape (bn, k)
# These appear to be probability distributions or softmax outputs
# so we'll make them positive and sum to 1 along dimension 1
x_out = torch.softmax(torch.randn(bn, k), dim=1)
x_tf_out = torch.softmax(torch.randn(bn, k), dim=1)

ARGS = {
    'x_out': x_out,
    'x_tf_out': x_tf_out
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (4, 4) — shape (4, 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
genclust Ran
ran on a synthesized input
nathanin/IIC/tensorflow/generate_clusterings.py
code served (permissive licence) · get_code("d17d2c856e1674fc")
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 scipy.special import softmax

np.random.seed(0)

# Parameter b: batch size (integer)
b = 2

# Parameter k: number of classes (integer)
k = 4

# Parameter y: optional one-hot encoded labels (numpy array or None)
# If provided, shape should be (b, k)
y = None

# Parameter seed: random seed (integer)
seed = 0

ARGS = {
    "b": b,
    "k": k,
    "y": y,
    "seed": seed,
}
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
mnist_x Ran
ran on a synthesized input
astirn/IIC/data.py
code served (permissive licence) · get_code("b8622cc70f86e3ba")
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 4, dtype family float, shape (2, 20, 20, 1) 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 tensorflow as tf

# Create a small MNIST-like image tensor
# Shape: (batch_size, height, width, channels)
batch_size = 2
height = 28
width = 28
channels = 1

# Create a random image tensor with uint8 dtype (typical for images)
x_orig = tf.constant(
    np.random.randint(0, 256, size=(batch_size, height, width, channels), dtype=np.uint8),
    dtype=tf.uint8
)

# Model input dimensions: (height, width, channels)
mdl_input_dims = (20, 20, 1)

# Boolean flag for training mode
is_training = True

ARGS = {
    "x_orig": x_orig,
    "mdl_input_dims": mdl_input_dims,
    "is_training": is_training,
}
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 4, dtype family float, shape (2, 20, 20, 1) — shape (2, 20, 20, 1) 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
pre_process_data Ran
ran on a synthesized input
astirn/IIC/data.py
code served (permissive licence) · get_code("7bd08b02a79cc9c7")
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 MockDataset. 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

class MockDatasetInfo:
    def __init__(self, name):
        self.name = name

class MockDataset:
    def __init__(self, num_samples=2):
        self.num_samples = num_samples
    
    def map(self, fn, num_parallel_calls=None):
        # Create mock data samples
        samples = []
        for i in range(self.num_samples):
            # Create a small MNIST-like image: batch_size=1, height=28, width=28, channels=1
            image = tf.constant(np.random.randint(0, 256, (28, 28, 1), dtype=np.uint8), dtype=tf.uint8)
            label = tf.constant(i % 10, dtype=tf.int64)
            sample = {'image': image, 'label': label}
            samples.append(sample)
        
        # Apply the function to each sample
        processed_samples = []
        for sample in samples:
            try:
                processed = fn(sample)
                processed_samples.append(processed)
            except Exception as e:
                # If processing fails, we still want to test the setup
                pass
        
        return self

ds = MockDataset(num_samples=2)
info = MockDatasetInfo(name='mnist')
is_training = True
mdl_input_dims = [28, 28, 1]
num_repeats = 2

ARGS = {
    'ds': ds,
    'info': info,
    'is_training': is_training,
    'mdl_input_dims': mdl_input_dims,
    'num_repeats': num_repeats
}
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 MockDataset — a MockDataset output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a MockDataset output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
unsupervised_labels Ran
ran on a synthesized input
astirn/IIC/utils.py
code served (permissive licence) · get_code("9ce81c4ed4d32382")
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.optimize import linear_sum_assignment

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

# y: true labels (ground truth class assignments)
# num_classes = 3, so labels should be in range [0, 2]
y = np.array([0, 1, 2, 0, 1, 2])

# y_hat: predicted cluster assignments
# num_clusters = 3, so predictions should be in range [0, 2]
y_hat = np.array([0, 1, 2, 0, 2, 1])

# num_classes and num_clusters must be equal
num_classes = 3
num_clusters = 3

ARGS = {
    'y': y,
    'y_hat': y_hat,
    'num_classes': num_classes,
    'num_clusters': num_clusters
}
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
convex_combo Not yet run
unverified (harvested, no recorded run)
nathanin/IIC/tensorflow/mnist_draw.py
code served (permissive licence) · get_code("7c744f563f3b9f6c")
How it ran

no invocation recorded.

convolution_layer Not yet run
unverified (harvested, no recorded run)
astirn/IIC/graphs.py
code served (permissive licence) · get_code("1d13a626952e713c")
How it ran

no invocation recorded.

fully_connected_layer Not yet run
unverified (harvested, no recorded run)
astirn/IIC/graphs.py
code served (permissive licence) · get_code("839dfd8e8303304d")
How it ran

no invocation recorded.

get_iterator Not yet run
unverified (harvested, no recorded run)
nathanin/IIC/tensorflow/data.py
code served (permissive licence) · get_code("95b736abab569c08")
How it ran

no invocation recorded.

get_subhead_using_loss Not yet run
unverified (harvested, no recorded run)
willi-menapace/acids-clustering-domain-shift/codebase/utils/cluster/cluster_eval.py
code served (permissive licence) · get_code("99bb4dfc8b575f7e")
How it ran

no invocation recorded.

max_pooling_layer Not yet run
unverified (harvested, no recorded run)
astirn/IIC/graphs.py
code served (permissive licence) · get_code("d73cf3ab592ee3f1")
How it ran

no invocation recorded.

mnist_gx Not yet run
unverified (harvested, no recorded run)
astirn/IIC/data.py
code served (permissive licence) · get_code("c7ef0b467f726ba3")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 18 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 17 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
compute_joint 1 other paper
1 later
2510.13887 (2025-10)
Incomplete Multi-view Clustering via Hierarchical Semantic Alignment and Cooperative Completion

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 present a novel clustering objective that learns a neural network classifier from scratch, given only unlabelled data samples. The model discovers clusters that accurately match semantic classes, achieving state-of-the-art results in eight unsupervised clustering benchmarks spanning image classification and segmentation. These include STL10, an unsupervised variant of ImageNet, and CIFAR10, where we significantly beat the accuracy of our closest competitors by 6.6 and 9.5 absolute percentage points respectively. The method is not specialised to computer vision and operates on any paired dataset samples; in our experiments we use random transforms to obtain a pair from each image. The trained network directly outputs semantic labels, rather than high dimensional representations that need external processing to be usable for semantic clustering. The objective is simply to maximise mutual information between the class assignments of each pair. It is easy to implement and rigorously grounded in information theory, meaning we effortlessly avoid degenerate solutions that other clustering methods are susceptible to. In addition to the fully unsupervised mode, we also test two semi-supervised settings. The first achieves 88.8% accuracy on STL10 classification, setting a new global state-of-the-art over all existing methods (whether supervised, semi-supervised or unsupervised). The second shows robustness to 90% reductions in label coverage, of relevance to applications that wish to make use of small amounts of labels. github.com/xu-ji/IIC

For agents

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

get_harvested_code_for_paper("1807.06653")
get_code_for_paper("1807.06653")
have("1807.06653")

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

curl https://syntology.ai/api/ran/1807.06653.json

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

[![Syntology run record](https://syntology.ai/api/ran/1807.06653.svg)](https://syntology.ai/paper/1807.06653)

Connect an agent — have() is free.