SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1905.05894 · 2019

Online Normalization for Training Neural Networks

arXiv · PDF · Open in the Atlas

Code that ran

9 of 16 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; 0 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/1905.05894.json · badge: /api/ran/1905.05894.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
cerebras/online-normalization canonical (classifier label)
classifier basis: census
9 ran · 7 no recorded run
FunctionStatusWhere it lives
add_square Ran
ran on a synthesized input
cerebras/online-normalization/experiments/unet/simulation.py
code served (permissive licence) · get_code("30240401a8068dfc")
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 (10, 10) 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 numpy as np

torch.manual_seed(0)

# Analyze the function:
# - arr: needs to support indexing with integers and slices, and assignment with boolean values
# - x, y: integers used for indexing
# - size: integer used in division and arithmetic for indexing

# Create a numpy array that can be indexed and modified as the function expects
arr = np.zeros((10, 10), dtype=bool)

x = 5
y = 5
size = 4

ARGS = {
    "arr": arr,
    "x": x,
    "y": y,
    "size": size
}
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 ndarray, ndim 2, dtype family float, shape (10, 10) — shape (10, 10) 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
generate_img_and_mask Ran
ran on a synthesized input
cerebras/online-normalization/experiments/unet/simulation.py
code served (permissive licence) · get_code("2840eb73965f941b")
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
import random

random.seed(0)
np.random.seed(0)

ARGS = {
    "height": 64,
    "width": 128
}
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
generate_random_data Ran
ran on a synthesized input
cerebras/online-normalization/experiments/unet/simulation.py
code served (permissive licence) · get_code("42517d21b4bbbe1d")
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 random
import numpy as np

random.seed(42)
np.random.seed(42)

ARGS = {
    "height": 64,
    "width": 64,
    "count": 2
}
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
get_parameter_groups Ran
ran on a synthesized input
cerebras/online-normalization/experiments/resnet/utils.py
code served (permissive licence) · get_code("d8bd6382fa212764")
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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create a minimal model with named parameters
class MinimalModel(nn.Module):
    def __init__(self):
        super().__init__()
        # Add various parameter types that the function looks for
        self.fc = nn.Linear(4, 3)  # 'fc' parameters should go to model_params
        self.norm1 = nn.BatchNorm1d(3)  # 'norm' parameters should go to norm_params
        self.bias_param = nn.Parameter(torch.randn(2))  # 'bias' in name should go to norm_params
        self.regular_param = nn.Parameter(torch.randn(5, 2))  # regular parameters go to model_params
    
    def forward(self, x):
        return x

model = MinimalModel()

# Ensure all parameters require gradients
for param in model.parameters():
    param.requires_grad = True

ARGS = {
    'model': model,
    'norm_weight_decay': 0.0001,
    'cifar': 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 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
mean_tensor Ran
ran on a synthesized input
cerebras/online-normalization/online-norm/pytorch/online_norm_pytorch/online_norm_2d.py
code served (permissive licence) · get_code("4994246847dd8bb1")
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 Tensor, ndim 3, dtype family float, shape (2, 1, 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 torch

torch.manual_seed(0)

# Create a small tensor: batch=2, dims=4 with shape (2, 4, 3)
input_tensor = torch.randn(2, 4, 3)

# norm_ax can be int or tuple - test with tuple
norm_ax_value = (1, 2)

# keepdim parameter
keepdim_value = True

ARGS = {
    "input": input_tensor,
    "norm_ax": norm_ax_value,
    "keepdim": keepdim_value
}
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 Tensor, ndim 3, dtype family float, shape (2, 1, 1) — shape (2, 1, 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
mult_scale_forward Ran
ran on a synthesized input
cerebras/online-normalization/online-norm/numpy_on/online_norm_1d.py
code served (permissive licence) · get_code("d29637ee5873d351")
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
import torch

torch.manual_seed(0)

# Analyze the function:
# - inputs: used with * operator and np.newaxis indexing, so needs to be array-like
# - weight: used with np.newaxis indexing and * operator, needs to be array-like
# Both are used in broadcasting: inputs * weight[np.newaxis, :]
# inputs should be shape (batch, features), weight should be shape (features,)

inputs = np.random.randn(2, 4)  # batch_size=2, feature_dim=4
weight = np.random.randn(4)      # feature_dim=4

ARGS = {
    "inputs": inputs,
    "weight": weight,
}
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 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
norm_backward Ran
ran on a synthesized input
cerebras/online-normalization/online-norm/numpy_on/online_norm_1d.py
code served (permissive licence) · get_code("bf2eb181871a83c7")
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

# Analyzing the function:
# - grad_out: numpy array that supports .shape[0] and indexing [idx]
# - ustream: numpy array used in arithmetic operations
# - vstream: numpy array used in arithmetic operations
# - abkw: scalar (float) used in arithmetic (1 - abkw)
# - cache: tuple containing (out, scale) where both are indexable numpy arrays

# Set up dimensions
batch_size = 2
feature_dim = 4

# Create grad_out: shape (batch_size, feature_dim)
grad_out = np.random.randn(batch_size, feature_dim).astype(np.float32)

# Create ustream: shape (feature_dim,) - accumulator for gradients
ustream = np.zeros(feature_dim, dtype=np.float32)

# Create vstream: shape (feature_dim,) - accumulator for gradients
vstream = np.zeros(feature_dim, dtype=np.float32)

# Create abkw: scalar float between 0 and 1 (appears to be a momentum/decay parameter)
abkw = 0.9

# Create cache: tuple of (out, scale)
# out: shape (batch_size, feature_dim) - output from forward pass
out = np.random.randn(batch_size, feature_dim).astype(np.float32)
# scale: shape (batch_size,) - normalization scale factor per sample
scale = np.random.randn(batch_size).astype(np.float32) + 1.0  # avoid near-zero
cache = (out, scale)

ARGS = {
    "grad_out": grad_out,
    "ustream": ustream,
    "vstream": vstream,
    "abkw": abkw,
    "cache": cache
}
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
norm_forward Ran
ran on a synthesized input
cerebras/online-normalization/online-norm/numpy_on/online_norm_1d.py
code served (permissive licence) · get_code("820691136676a1ce")
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 seed for reproducibility
np.random.seed(0)

# Create small tensors with distinct shapes
batch_size = 2
feature_dim = 4

# inputs: batch of samples, shape (batch_size, feature_dim)
inputs = np.random.randn(batch_size, feature_dim).astype(np.float32)

# mstream: running mean, shape (feature_dim,)
mstream = np.random.randn(feature_dim).astype(np.float32)

# varstream: running variance, shape (feature_dim,)
varstream = np.abs(np.random.randn(feature_dim).astype(np.float32)) + 0.1

# afwd: decay factor (scalar, typically between 0 and 1)
afwd = 0.9

# eps: small epsilon for numerical stability
eps = 1e-5

ARGS = {
    "inputs": inputs,
    "mstream": mstream,
    "varstream": varstream,
    "afwd": afwd,
    "eps": eps,
}
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
norm_forward Ran
ran on a synthesized input
cerebras/online-normalization/online-norm/numpy_on/online_norm_2d.py
code served (permissive licence) · get_code("60e3e0091d896b87")
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 seed for reproducibility
np.random.seed(0)

# Create tiny test inputs
batch_size = 2
num_channels = 4
height = 3
width = 5

# inputs: shape (batch, channels, height, width)
inputs = np.random.randn(batch_size, num_channels, height, width).astype(np.float32)

# mstream: running mean, shape (channels,)
mstream = np.zeros(num_channels, dtype=np.float32)

# varstream: running variance, shape (channels,)
varstream = np.ones(num_channels, dtype=np.float32)

# afwd: momentum/blend factor for exponential moving average
afwd = 0.9

# eps: small constant for numerical stability
eps = 1e-5

ARGS = {
    "inputs": inputs,
    "mstream": mstream,
    "varstream": varstream,
    "afwd": afwd,
    "eps": eps
}
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
accuracy Not yet run
unverified (harvested, no recorded run)
cerebras/online-normalization/experiments/resnet/utils.py
code served (permissive licence) · get_code("23d0bd9f1b241feb")
How it ran

no invocation recorded.

conv_alongb_w1 Not yet run
unverified (harvested, no recorded run)
cerebras/online-normalization/online-norm/pytorch/online_norm_pytorch/online_norm_1d.py
code served (permissive licence) · get_code("50f69e4173cfbf3b")
How it ran

no invocation recorded.

conv_alongb_w1 Not yet run
unverified (harvested, no recorded run)
cerebras/online-normalization/online-norm/pytorch/online_norm_pytorch/online_norm_2d.py
code served (permissive licence) · get_code("cf4866a0e5413a8e")
How it ran

no invocation recorded.

lin_crtl Not yet run
unverified (harvested, no recorded run)
cerebras/online-normalization/online-norm/pytorch/online_norm_pytorch/online_norm_1d.py
code served (permissive licence) · get_code("c3c01085619cd9fa")
How it ran

no invocation recorded.

lin_momentum Not yet run
unverified (harvested, no recorded run)
cerebras/online-normalization/online-norm/pytorch/online_norm_pytorch/online_norm_1d.py
code served (permissive licence) · get_code("1ab49f7983db64d0")
How it ran

no invocation recorded.

lin_momentum Not yet run
unverified (harvested, no recorded run)
cerebras/online-normalization/online-norm/pytorch/online_norm_pytorch/online_norm_2d.py
code served (permissive licence) · get_code("846d691e31cde436")
How it ran

no invocation recorded.

validate Not yet run
unverified (harvested, no recorded run)
cerebras/online-normalization/experiments/resnet/utils.py
code served (permissive licence) · get_code("80a8d95be697182f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 1 the same month, 1 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
accuracy 80 other papers
18 dated earlier, 2 the same month, 60 later
1307.0414 (2013-07)
Challenges in Representation Learning: A report on three machine learning contests
1412.6553 (2014-12)
Speeding-up Convolutional Neural Networks Using Fine-tuned CP-Decomposition
1503.02531 (2015-03)
Distilling the Knowledge in a Neural Network
and 77 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

Online Normalization is a new technique for normalizing the hidden activations of a neural network. Like Batch Normalization, it normalizes the sample dimension. While Online Normalization does not use batches, it is as accurate as Batch Normalization. We resolve a theoretical limitation of Batch Normalization by introducing an unbiased technique for computing the gradient of normalized activations. Online Normalization works with automatic differentiation by adding statistical normalization as a primitive. This technique can be used in cases not covered by some other normalizers, such as recurrent networks, fully connected networks, and networks with activation memory requirements prohibitive for batching. We show its applications to image classification, image segmentation, and language modeling. We present formal proofs and experimental results on ImageNet, CIFAR, and PTB datasets.

For agents

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

get_harvested_code_for_paper("1905.05894")
get_code_for_paper("1905.05894")
have("1905.05894")

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

curl https://syntology.ai/api/ran/1905.05894.json

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

[![Syntology run record](https://syntology.ai/api/ran/1905.05894.svg)](https://syntology.ai/paper/1905.05894)

Connect an agent — have() is free.