SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1502.03167 · ICML · 2015

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

Sergey Ioffe

arXiv · PDF · Open in the Atlas

Code that ran

19 of 21 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (5 of them constructed an object rather than computing a result; 2 honoured the contract we drafted, 0 violated it and 10 have no contract verdict; 7 failures of our instrument; 2 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/1502.03167.json · badge: /api/ran/1502.03167.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
yangyucheng000/ssd_inception_v2 community (classifier label)
classifier basis not recorded on the link
5 ran · 1 no recorded run
labmlai/annotated_deep_learning_paper_implementations community
not in the paper's declared repositories
2 ran · 0 no recorded run
dichotomies/proxy-nca community
not in the paper's declared repositories
1 ran · 0 no recorded run
zjZSTU/GoogLeNet community
not in the paper's declared repositories
1 ran · 0 no recorded run
tanjeffreyz/batch-normalization community
not in the paper's declared repositories
1 ran · 0 no recorded run
yzheng51/rl-dino-run community
not in the paper's declared repositories
1 ran · 0 no recorded run
Sakib1263/Inception-Model-Builder-Tensorflow-Keras community
Papers with Code archive
1 ran · 0 no recorded run
Sakib1263/ResNet-ResNeXt-1D-2D-Tensorflow-Keras community
Papers with Code archive
1 ran · 0 no recorded run
Sakib1263/VGG-1D-2D-Tensorflow-Keras community
Papers with Code archive
1 ran · 0 no recorded run
Sakib1263/VGG-Model-Builder-KERAS community
Papers with Code archive
1 ran · 0 no recorded run
tensorflow/models community
not in the paper's declared repositories
1 ran · 0 no recorded run
Shantanu48114860/Handwritten-Character-Recognition community
Papers with Code archive
1 ran · 0 no recorded run
minoring/batch-norm-visualize community
not in the paper's declared repositories
1 ran · 0 no recorded run
karurb92/ldam_str_bn community
not in the paper's declared repositories
1 ran · 0 no recorded run
sayakpaul/Adaptive-Gradient-Clipping community
not in the paper's declared repositories
0 ran · 1 no recorded run
FunctionStatusWhere it lives
BNInception Ran
ran on a synthesized input · constructed an object; no result was computed
dichotomies/proxy-nca/net/bn_inception.py
code served (permissive licence) · get_code("dcefb760e5798c17")
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 an instance of its own class, BNInception: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn

torch.manual_seed(0)

ARGS = {
    "nb_classes": 1000,
    "is_pretrained": False,
}
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 BNInception — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a BNInception 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
BasicConv2d Ran
ran on a synthesized input
zjZSTU/GoogLeNet/py/models/googlenet_bn.py
code served (permissive licence) · get_code("363095e135affba2")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 4, dtype family float, shape (2, 4, 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 torch
import torch.nn as nn

class BasicConv2d(nn.Module):

    def __init__(self, in_channels, out_channels, **kwargs):
        super(BasicConv2d, self).__init__()
        self.conv = nn.Conv2d(in_channels, out_channels, bias=False, **kwargs)
        self.bn = nn.BatchNorm2d(out_channels, eps=0.001)

    def forward(self, x):
        x = self.conv(x)
        return torch.nn.functional.relu(x, inplace=True)


INIT_ARGS = {
    'in_channels': 3,
    'out_channels': 4,
    'kernel_size': 3,
    'padding': 1,
}

ARGS = {
    'x': torch.randn(2, 3, 8, 8),
}

if __name__ == '__main__':
    model = BasicConv2d(**INIT_ARGS)
    output = model(**ARGS)
    print(f"Output shape: {output.shape}")
    print(f"Expected shape: torch.Size([2, 4, 8, 8])")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 4, dtype family float, shape (2, 4, 8, 8) — shape (2, 4, 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
BatchNorm Ran
ran on a synthesized input · constructed an object; no result was computed
labmlai/annotated_deep_learning_paper_implementations/labml_nn/normalization/batch_norm/mnist.py
code served (permissive licence) · get_code("629771a2e32edf83")
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 an instance of its own class, BatchNorm: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'channels': 4,
    'eps': 1e-05,
    'momentum': 0.1,
    'affine': True,
    'track_running_stats': True,
}
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 BatchNorm — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a BatchNorm 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
BatchNorm Ran
ran on a synthesized input
tanjeffreyz/batch-normalization/models.py
pointer only (licence: NONE) · get_code("fd8f4df6039d46b0")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, tanjeffreyz/batch-normalization when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

ConvNet Ran
ran on a synthesized input
yzheng51/rl-dino-run/model.py
code served (permissive licence) · get_code("275928042f41cfff")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 2, dtype family float, shape (2, 3136) 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
import torch.nn as nn
import torch.nn.functional as F

class ConvNet(nn.Module):
    """Convolutional layers in DQN"""

    def __init__(self, batch_norm=False):
        super().__init__()
        self.conv1 = nn.Conv2d(4, 32, kernel_size=8, stride=4)
        self.conv2 = nn.Conv2d(32, 64, kernel_size=4, stride=2)
        self.conv3 = nn.Conv2d(64, 64, kernel_size=3, stride=1)
        if batch_norm:
            self.bn1 = nn.BatchNorm2d(32)
            self.bn2 = nn.BatchNorm2d(64)
            self.bn3 = nn.BatchNorm2d(64)
        else:
            self.bn1 = nn.Identity()
            self.bn2 = nn.Identity()
            self.bn3 = nn.Identity()

    def forward(self, x):
        x = F.relu(self.bn1(self.conv1(x)))
        x = F.relu(self.bn2(self.conv2(x)))
        x = F.relu(self.bn3(self.conv3(x)))
        return x.view(x.size(0), -1)


INIT_ARGS = {
    'batch_norm': False
}

ARGS = {
    'x': torch.randn(2, 4, 84, 84)
}

if __name__ == '__main__':
    model = ConvNet(**INIT_ARGS)
    output = model(**ARGS)
    print(f"Output shape: {output.shape}")
    print(f"Output: {output}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 (2, 3136) — shape (2, 3136) 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
Conv_1D_Block Ran
ran; our contract draft was wrong, not the code
Sakib1263/Inception-Model-Builder-Tensorflow-Keras/Codes/Inception_ResNet_1DCNN.py
code served (permissive licence) · get_code("977f839ae1b772a6")
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) and returned KerasTensor. 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 a small input tensor for Conv1D
# Shape: (batch_size, sequence_length, input_channels)
# Using batch=2, sequence_length=4, input_channels=3
x = tf.keras.Input(shape=(4, 3))

# Define the arguments dictionary
ARGS = {
    'x': x,
    'model_width': 8,
    'kernel': 3,
    'strides': 1,
    'padding': 'same'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type KerasTensor — a KerasTensor 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
Conv_1D_Block Ran
ran; our contract draft was wrong, not the code
Sakib1263/ResNet-ResNeXt-1D-2D-Tensorflow-Keras/Codes/ResNet_1DCNN.py
code served (permissive licence) · get_code("fa041da57fc2fe89")
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 KerasTensor. 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 a small input tensor
# Shape: (batch_size, sequence_length, input_channels)
# Using distinct dimensions: batch=2, seq_len=4, channels=3
x = tf.keras.Input(shape=(4, 3), batch_size=2)

# Define the parameters
model_width = 5  # output channels
kernel = 3  # kernel size
strides = 1  # stride

ARGS = {
    'x': x,
    'model_width': model_width,
    'kernel': kernel,
    'strides': strides
}
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 KerasTensor — a KerasTensor 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
Conv_1D_Block Ran
ran; our contract draft was wrong, not the code
Sakib1263/VGG-1D-2D-Tensorflow-Keras/Codes/VGG_1DCNN.py
code served (permissive licence) · get_code("4f854d75a8d26231")
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) and returned KerasTensor. 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 a small input tensor
# Shape: (batch_size=2, sequence_length=4, input_channels=3)
x = tf.keras.Input(shape=(4, 3))

# Define parameters
model_width = 8  # number of filters in Conv1D
kernel = 3       # kernel size

# Create the ARGS dictionary
ARGS = {
    'x': x,
    'model_width': model_width,
    'kernel': kernel
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type KerasTensor — a KerasTensor 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
Conv_2D_Block Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Sakib1263/VGG-Model-Builder-KERAS/Codes/VGG_2DCNN.py
code served (permissive licence) · get_code("12a688d4588443d2")
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) and returned EagerTensor, ndim 4, dtype family float, shape (2, 4, 4, 16) 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

# Create a small input tensor (batch_size=2, height=4, width=4, channels=3)
inputs = tf.keras.Input(shape=(4, 4, 3))

# Create a simple model to get a concrete tensor
model = tf.keras.Sequential([
    tf.keras.layers.Conv2D(3, (1, 1), padding='same', input_shape=(4, 4, 3))
])

# Generate a concrete input tensor
concrete_inputs = tf.random.normal((2, 4, 4, 3))

# Define arguments
ARGS = {
    'inputs': concrete_inputs,
    'model_width': 16,
    'kernel': 3
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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, 4, 4, 16) — shape (2, 4, 4, 16) 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
FreezableBatchNorm Ran
ran on a synthesized input · constructed an object; no result was computed
tensorflow/models/research/object_detection/core/freezable_batch_norm.py
pointer only (licence: NOASSERTION) · get_code("d6503cf479ae48ea")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, tensorflow/models (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

Model Ran
ran on a synthesized input · constructed an object; no result was computed
labmlai/annotated_deep_learning_paper_implementations/labml_nn/normalization/batch_norm/mnist.py
code served (permissive licence) · get_code("84ac6eb3495fd518")
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 an instance of its own class, Model: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn
import torch.nn.functional as F

torch.manual_seed(0)

# The Model class takes no arguments in __init__
ARGS = {}
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 Model — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Model 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
Network Ran
ran on a synthesized input · constructed an object; no result was computed
Shantanu48114860/Handwritten-Character-Recognition/Handwritten-Character-Recognition/CNN.py
pointer only (licence: NONE) · get_code("2ba90697099f8cca")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Shantanu48114860/Handwritten-Character-Recognition when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

apply_nms Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yangyucheng000/ssd_inception_v2/src/eval_utils.py
code served (permissive licence) · get_code("9a71632baa853b8a")
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 int; 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

# Create sample bounding boxes in format [y1, x1, y2, x2]
all_boxes = np.array([
    [10.0, 20.0, 50.0, 60.0],
    [12.0, 22.0, 52.0, 62.0],
    [100.0, 120.0, 150.0, 160.0],
    [200.0, 220.0, 250.0, 260.0]
], dtype=np.float32)

# Create corresponding scores
all_scores = np.array([0.9, 0.8, 0.7, 0.6], dtype=np.float32)

# NMS threshold for overlap
thres = 0.3

# Maximum number of boxes to keep
max_boxes = 3

ARGS = {
    "all_boxes": all_boxes,
    "all_scores": all_scores,
    "thres": thres,
    "max_boxes": max_boxes
}
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 list, ndim 1, dtype family int — 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
get_depth_fn Ran
ran on a synthesized input
yangyucheng000/ssd_inception_v2/src/feature_map_generators.py
code served (permissive licence) · get_code("0946825bd8511f99")
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 function. 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 numpy as np

torch.manual_seed(0)

ARGS = {
    "depth_multipler": 0.5,
    "min_depth": 8,
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
get_imageId_from_fileName Ran
ran, honoured the contract we drafted
yangyucheng000/ssd_inception_v2/src/dataset.py
code served (permissive licence) · get_code("80eae5f06649b1e2")
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 int, ndim 0, dtype family int, 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 os

ARGS = {
    "filename": "12345.jpg",
    "id_iter": 42
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_lr Ran
ran, honoured the contract we drafted
yangyucheng000/ssd_inception_v2/src/lr_schedule.py
code served (permissive licence) · get_code("1bab04258df1af4e")
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 (990,) 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 math
import numpy as np

ARGS = {
    'global_step': 10,
    'lr_init': 0.0001,
    'lr_end': 0.00001,
    'lr_max': 0.001,
    'warmup_epochs': 2,
    'total_epochs': 10,
    'steps_per_epoch': 100,
}
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 (990,) — shape (990,) 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
intersect Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yangyucheng000/ssd_inception_v2/src/box_utils.py
code served (permissive licence) · get_code("f4aebce5f37270c3")
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

# Construct valid arguments for the intersect function
# The function computes intersection of bounding boxes
# box_a should be shape (N, 4) - multiple boxes
# box_b should be shape (4,) - single box
# Each box is [y_min, x_min, y_max, x_max]

box_a = np.array([
    [10.0, 20.0, 30.0, 40.0],
    [15.0, 25.0, 35.0, 45.0]
], dtype=np.float32)

box_b = np.array([12.0, 22.0, 32.0, 42.0], dtype=np.float32)

ARGS = {
    "box_a": box_a,
    "box_b": box_b
}
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
mnist_model_bn Ran
ran on a synthesized input
minoring/batch-norm-visualize/model.py
pointer only (licence: NONE) · get_code("17bcce71dd6805e5")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, minoring/batch-norm-visualize when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

res_net_block Ran
ran; our contract draft was wrong, not the code
karurb92/ldam_str_bn/models/resnet.py
pointer only (licence: NONE) · get_code("a542810366586242")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, karurb92/ldam_str_bn when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

conv_block_bn Not yet run
unverified (harvested, no recorded run)
sayakpaul/Adaptive-Gradient-Clipping/models.py
code served (permissive licence) · get_code("b68ffa76fbfa3320")
How it ran

no invocation recorded.

random_sample_crop Not yet run
unverified (harvested, no recorded run)
yangyucheng000/ssd_inception_v2/src/dataset.py
code served (permissive licence) · get_code("f50e9b77fc01e9c8")
How it ran

no invocation recorded.

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, 7 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 7 later) and 14 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
Conv_1D_Block
brick of this paper's assembly: entry
3 other papers
3 later · 1 only as an assembly brick
1512.03385 (2015-12)
Deep Residual Learning for Image Recognition
1608.06993 (2016-08)
Densely Connected Convolutional Networks
1611.05431 (2016-11)
Aggregated Residual Transformations for Deep Neural Networks
intersect 2 other papers
2 later
1512.02325 (2015-12)
SSD: Single Shot MultiBox Detector
2006.04388 (2020-06)
Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection
random_sample_crop 2 other papers
2 later
1512.02325 (2015-12)
SSD: Single Shot MultiBox Detector
2006.04388 (2020-06)
Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection
ConvNet
brick of this paper's assembly: entry
1 other paper
1 later
1509.06461 (2015-09)
Deep Reinforcement Learning with Double Q-learning
apply_nms 1 other paper
1 later
1512.02325 (2015-12)
SSD: Single Shot MultiBox Detector
get_imageId_from_fileName 1 other paper
1 later
1512.02325 (2015-12)
SSD: Single Shot MultiBox Detector
get_lr 1 other paper
1 later
1905.02244 (2019-05)
Searching for MobileNetV3

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

More are linked than shown here.

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

Abstract

Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization. It also acts as a regularizer, in some cases eliminating the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batchnormalized networks, we improve upon the best published result on ImageNet classification: reaching 4.9% top-5 validation error (and 4.8% test error), exceeding the accuracy of human raters.

For agents

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

get_harvested_code_for_paper("1502.03167")
get_code_for_paper("1502.03167")
have("1502.03167")

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

curl https://syntology.ai/api/ran/1502.03167.json

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

[![Syntology run record](https://syntology.ai/api/ran/1502.03167.svg)](https://syntology.ai/paper/1502.03167)

Connect an agent — have() is free.