SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1911.06987 · 2019

Faster AutoAugment: Learning Augmentation Strategies using Backpropagation

arXiv · PDF · Open in the Atlas

Code that ran

6 of 9 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 5 have no contract verdict; 1 failure of our instrument; 3 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/1911.06987.json · badge: /api/ran/1911.06987.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
moskomule/dda community (classifier label)
classifier basis not recorded on the link
6 ran · 3 no recorded run
FunctionStatusWhere it lives
get_gaussian_3x3kernel Ran
ran on a synthesized input
moskomule/dda/dda/kernels.py
code served (permissive licence) · get_code("1ccd9eba5721f000")
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 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 torch

torch.manual_seed(0)

ARGS = {
    "sigma": torch.tensor([0.5, 1.0, 1.5, 2.0]),
    "device": None,
}
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 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
get_gaussian_5x5kernel Ran
ran on a synthesized input
moskomule/dda/dda/kernels.py
code served (permissive licence) · get_code("c74fe1c3169c6c7e")
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 2, dtype family float, shape (5, 5) 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)

# The function get_gaussian_5x5kernel takes:
# - sigma: torch.Tensor (used in _gaussian where .mean() and .pow(2) are called)
# - device: Optional[torch.device] (optional, defaults to None)

# Create a small sigma tensor
sigma = torch.tensor([1.5, 2.0, 1.2, 0.8])

# Device is optional, we'll pass None to use CPU
device = None

ARGS = {
    "sigma": sigma,
    "device": device,
}
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 2, dtype family float, shape (5, 5) — shape (5, 5) 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_sharpness_kernel Ran
ran on a synthesized input
moskomule/dda/dda/kernels.py
code served (permissive licence) · get_code("168e865c2a5bfb4d")
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 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 torch

torch.manual_seed(0)

ARGS = {
    "device": None
}
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 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
pil_function Ran
ran on a synthesized input
moskomule/dda/dda/pil.py
code served (permissive licence) · get_code("b4369762852f94d3")
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.

from PIL import Image
import functools

def pil_function(func):
    @functools.wraps(func)
    def inner(*args):
        if not isinstance(args[0], PILImage):
            raise RuntimeError(f'img is expected to be PIL.Image, but got {type(args[0])} instead!')
        return func(*args)
    return inner

# Create a sample PIL Image
sample_image = Image.new('RGB', (64, 64), color=(255, 0, 0))

# Define a simple function to be decorated
@pil_function
def process_image(img):
    return img.size

ARGS = {
    'func': process_image.__wrapped__
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL — 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
ste Ran
ran on a synthesized input
moskomule/dda/dda/functional.py
code served (permissive licence) · get_code("b67da72b4fb68fb1")
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 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 torch

torch.manual_seed(0)

# Create two tensors with different shapes to test the STE function
# input_forward will be returned as-is in the forward pass
# input_backward will be used to store shape information for the backward pass
input_forward = torch.randn(2, 4, requires_grad=True)
input_backward = torch.randn(2, 3, 4, requires_grad=True)

ARGS = {
    "input_forward": input_forward,
    "input_backward": input_backward,
}
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 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
tensor_function Ran
ran; our contract draft was wrong, not the code
moskomule/dda/dda/functional.py
code served (permissive licence) · get_code("f96aedd60a8e532e")
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 functools
import torch
import torch.nn as nn

torch.manual_seed(0)

# Define a minimal decorated function that works with the decorator
@functools.wraps(lambda *args: None)
def dummy_func(img, mag, kernel=None):
    """Dummy function that the decorator will wrap"""
    if kernel is not None:
        return img * 0.5
    elif mag is not None:
        if torch.is_tensor(mag):
            return img * mag
        return img * 0.8
    return img * 0.9

# Apply the decorator
def tensor_function(func):
    @functools.wraps(func)
    def inner(*args):
        if len(args) == 1:
            img = args[0]
            mag = None
        elif len(args) == 2:
            img, mag = args
        else:
            img, mag, kernel = args
        if not torch.is_tensor(img):
            raise RuntimeError(f'img is expected to be torch.Tensor, but got {type(img)} instead')
        if img.dim() == 3:
            img = img.unsqueeze(0)
        if torch.is_tensor(mag) and mag.nelement() != 1 and (mag.size(0) != img.size(0)):
            raise RuntimeError('Shape of `mag` is expected to be `B`')
        out = func(img, mag, kernel) if len(args) == 3 else func(img, mag)
        return out.clamp_(0, 1)
    return inner

decorated_func = tensor_function(dummy_func)

# Create test arguments
img_2d = torch.rand(4, 3)  # 3D tensor: will be unsqueezed to (1, 4, 3)
img_batch = torch.rand(2, 4, 3)  # 4D tensor: batch of 2
mag_scalar = torch.tensor(0.5)  # scalar magnitude
mag_batch = torch.rand(2)  # batch magnitude matching batch size
kernel = torch.rand(3, 3)  # simple kernel

ARGS = {
    'func': decorated_func
}
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 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
shear_x Not yet run
unverified (harvested, no recorded run)
moskomule/dda/dda/functional.py
code served (permissive licence) · get_code("90f7118023305b3b")
How it ran

no invocation recorded.

shear_x Not yet run
unverified (harvested, no recorded run)
moskomule/dda/dda/pil.py
code served (permissive licence) · get_code("1bcc85675f7606d4")
How it ran

no invocation recorded.

shear_y Not yet run
unverified (harvested, no recorded run)
moskomule/dda/dda/pil.py
code served (permissive licence) · get_code("8959fa6a39629fbc")
How it ran

no invocation recorded.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 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 8 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
tensor_function 1 other paper
1 later
2401.10474 (2024-01)
LDReg: Local Dimensionality Regularized Self-Supervised Learning

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

Data augmentation methods are indispensable heuristics to boost the performance of deep neural networks, especially in image recognition tasks. Recently, several studies have shown that augmentation strategies found by search algorithms outperform hand-made strategies. Such methods employ black-box search algorithms over image transformations with continuous or discrete parameters and require a long time to obtain better strategies. In this paper, we propose a differentiable policy search pipeline for data augmentation, which is much faster than previous methods. We introduce approximate gradients for several transformation operations with discrete parameters as well as the differentiable mechanism for selecting operations. As the objective of training, we minimize the distance between the distributions of augmented data and the original data, which can be differentiated. We show that our method, Faster AutoAugment, achieves significantly faster searching than prior work without a performance drop.

For agents

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

get_harvested_code_for_paper("1911.06987")
get_code_for_paper("1911.06987")
have("1911.06987")

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

curl https://syntology.ai/api/ran/1911.06987.json

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

[![Syntology run record](https://syntology.ai/api/ran/1911.06987.svg)](https://syntology.ai/paper/1911.06987)

Connect an agent — have() is free.