SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2003.03780 · 2020

DADA: Differentiable Automatic Data Augmentation

arXiv · PDF · Open in the Atlas

Code that ran

8 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; 3 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 5 failures of our instrument; 1 no recorded run), measured 2026-11-21; 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/2003.03780.json · badge: /api/ran/2003.03780.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
VDIGPKU/DADA canonical (classifier label)
classifier basis: census
8 ran · 1 no recorded run
FunctionStatusWhere it lives
conv3x3 Ran
ran; our contract draft was wrong, not the code
VDIGPKU/DADA/search_gumbel/networks/resnet.py
pointer only (licence: MIT); the stored copy get_code serves, from lixiaotong97/DSU, is pointer only (licence: NONE) · get_code("fac5364e2f53c6db")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from lixiaotong97/DSU, and GitHub detected no licence in that repository 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.

conv3x3 Ran
ran; our contract draft was wrong, not the code
VDIGPKU/DADA/search_gumbel/networks/wideresnet.py
pointer only (licence: MIT); the stored copy get_code serves, from DaikiTanak/manifold_mixup, is pointer only (licence: NONE) · get_code("00e569acd6b45ef0")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from DaikiTanak/manifold_mixup, and GitHub detected no licence in that repository 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.

conv3x3 Ran
ran; our contract draft was wrong, not the code
VDIGPKU/DADA/search_gumbel/networks/pyramidnet.py
pointer only (licence: MIT); the stored copy get_code serves, from lzj994/Binary-Quantization, is pointer only (licence: NONE) · get_code("03a0be4fb7381bd2")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from lzj994/Binary-Quantization, and GitHub detected no licence in that repository 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.

float_parameter Ran
ran, honoured the contract we drafted
VDIGPKU/DADA/fast-autoaugment/archive.py
pointer only (licence: MIT); the stored copy get_code serves, from chenchenzong/dpc, is pointer only (licence: NONE) · get_code("40d3c39a403ae1f7")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from chenchenzong/dpc, and GitHub detected no licence in that repository 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.

int_parameter Ran
ran, honoured the contract we drafted
VDIGPKU/DADA/fast-autoaugment/archive.py
pointer only (licence: MIT); the stored copy get_code serves, from chenchenzong/dpc, is pointer only (licence: NONE) · get_code("7900a3a6deef908b")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from chenchenzong/dpc, and GitHub detected no licence in that repository 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.

parse_devkit Ran
ran; our contract draft was wrong, not the code
VDIGPKU/DADA/search_gumbel/imagenet.py
code served (permissive licence) · get_code("10d196b211843339")
How it ran

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

import os
import tempfile
import scipy.io as sio
import numpy as np

# Create a temporary directory structure with necessary files
tmpdir = tempfile.mkdtemp()
data_dir = os.path.join(tmpdir, 'data')
os.makedirs(data_dir, exist_ok=True)

# Create a minimal meta.mat file
# The meta structure needs synsets with fields indexed at position 4 being num_children
# Structure: (idx, wnid, classes, parent_idx, num_children, ...)
meta_data = np.array([
    (1, 'n0001', 'dog, puppy', 0, 0),
    (2, 'n0002', 'cat, feline', 0, 0),
    (3, 'n0003', 'animal', 0, 2),
], dtype=[('idx', 'i4'), ('wnid', 'U10'), ('classes', 'U50'), ('parent', 'i4'), ('num_children', 'i4')])

sio.savemat(
    os.path.join(data_dir, 'meta.mat'),
    {'synsets': meta_data},
    do_compression=False
)

# Create validation ground truth file
val_ground_truth_path = os.path.join(data_dir, 'ILSVRC2012_validation_ground_truth.txt')
with open(val_ground_truth_path, 'w') as f:
    f.write('1\n')
    f.write('2\n')
    f.write('1\n')

ARGS = {
    'root': tmpdir
}
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 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
parse_meta Ran
ran; our contract draft was wrong, not the code
VDIGPKU/DADA/search_gumbel/imagenet.py
code served (permissive licence) · get_code("97bfaa2cce837d67")
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 os
import tempfile
import scipy.io as sio
import numpy as np

# Create a temporary directory and meta.mat file
temp_dir = tempfile.mkdtemp()
data_dir = os.path.join(temp_dir, 'data')
os.makedirs(data_dir, exist_ok=True)

# Create synthetic synsets data matching the expected structure
# Each synset entry: (idx, wnid, class_name, something, num_children, ...)
# We need num_children == 0 for filtering
synsets = np.array([
    (0, 'n00001', 'animal, creature', 'def1', 0),
    (1, 'n00002', 'plant, flora', 'def2', 2),  # num_children=2, will be filtered out
    (2, 'n00003', 'car, automobile', 'def3', 0),
    (3, 'n00004', 'dog, canine', 'def4', 1),   # num_children=1, will be filtered out
], dtype=[('idx', 'i4'), ('wnid', 'U10'), ('class', 'U30'), ('def', 'U20'), ('num_children', 'i4')])

# Save as meta.mat
meta_file = os.path.join(data_dir, 'meta.mat')
sio.savemat(meta_file, {'synsets': synsets})

ARGS = {
    'devkit_root': temp_dir,
    'path': 'data',
    'filename': 'meta.mat'
}
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 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
parse_val_groundtruth Ran
ran, honoured the contract we drafted
VDIGPKU/DADA/search_gumbel/imagenet.py
code served (permissive licence) · get_code("e7825a1705689714")
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 os
import tempfile

# Create a temporary directory and validation ground truth file
temp_dir = tempfile.mkdtemp()
data_dir = os.path.join(temp_dir, 'data')
os.makedirs(data_dir, exist_ok=True)

# Create a sample validation ground truth file with class indices
val_gt_content = "1\n5\n10\n2\n7\n"
gt_file = os.path.join(data_dir, 'ILSVRC2012_validation_ground_truth.txt')
with open(gt_file, 'w') as f:
    f.write(val_gt_content)

ARGS = {
    'devkit_root': temp_dir,
    'path': 'data',
    'filename': 'ILSVRC2012_validation_ground_truth.txt'
}
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 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
num_class Not yet run
unverified (harvested, no recorded run)
VDIGPKU/DADA/search_gumbel/dataset.py
code served (permissive licence) · get_code("ba7d4c39d8ffaaa2")
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 8 distinct functions with an identity, 7 are also present in other papers whose code we hold (7 with at least one paper dated earlier, 2 the same month, 7 later) and 1 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
conv3x3 686 other papers
197 dated earlier, 17 the same month, 471 later, 1 same year, order not determinable · 31 only as an assembly brick
1406.2661 (2014-06)
Generative Adversarial Networks
1409.0575 (2014-09)
ImageNet Large Scale Visual Recognition Challenge
1412.6572 (2014-12)
Explaining and Harnessing Adversarial Examples
and 683 more not named here
conv3x3 74 other papers
14 dated earlier, 1 the same month, 59 later · 5 only as an assembly brick
1605.07146 (2016-05)
Wide Residual Networks
1707.01629 (2017-07)
Dual Path Networks
1802.05668 (2018-02)
Model compression via distillation and quantization
and 71 more not named here
float_parameter 15 other papers
7 dated earlier, 8 later
1409.4842 (2014-09)
Going deeper with convolutions
1512.00567 (2015-12)
Rethinking the Inception Architecture for Computer Vision
1608.06993 (2016-08)
Densely Connected Convolutional Networks
and 12 more not named here
int_parameter 6 other papers
3 dated earlier, 3 later
1904.12848 (2019-04)
Unsupervised Data Augmentation for Consistency Training
1905.00397 (2019-05)
Fast AutoAugment
2001.06001 (2020-01)
Curriculum Labeling: Revisiting Pseudo-Labeling for Semi-Supervised Learning
and 3 more not named here
parse_devkit 4 other papers
2 dated earlier, 2 later
1904.01569 (2019-04)
Exploring Randomly Wired Neural Networks for Image Recognition
1905.00397 (2019-05)
Fast AutoAugment
2006.05553 (2020-06)
Neural Methods for Point-wise Dependency Estimation
and 1 more not named here
parse_meta 3 other papers
1 dated earlier, 2 later
1905.00397 (2019-05)
Fast AutoAugment
2006.05553 (2020-06)
Neural Methods for Point-wise Dependency Estimation
parse_val_groundtruth 3 other papers
1 dated earlier, 2 later
1905.00397 (2019-05)
Fast AutoAugment
2006.05553 (2020-06)
Neural Methods for Point-wise Dependency Estimation

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 2026-11-21.

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 (DA) techniques aim to increase data variability, and thus train deep networks with better generalisation. The pioneering AutoAugment automated the search for optimal DA policies with reinforcement learning. However, AutoAugment is extremely computationally expensive, limiting its wide applicability. Followup works such as Population Based Augmentation (PBA) and Fast AutoAugment improved efficiency, but their optimization speed remains a bottleneck. In this paper, we propose Differentiable Automatic Data Augmentation (DADA) which dramatically reduces the cost. DADA relaxes the discrete DA policy selection to a differentiable optimization problem via Gumbel-Softmax. In addition, we introduce an unbiased gradient estimator, RELAX, leading to an efficient and effective one-pass optimization strategy to learn an efficient and accurate DA policy. We conduct extensive experiments on CIFAR-10, CIFAR-100, SVHN, and ImageNet datasets. Furthermore, we demonstrate the value of Auto DA in pre-training for downstream detection problems. Results show our DADA is at least one order of magnitude faster than the state-of-the-art while achieving very comparable accuracy. The code is available at https://github.com/VDIGPKU/DADA.

For agents

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

get_harvested_code_for_paper("2003.03780")
get_code_for_paper("2003.03780")
have("2003.03780")

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

curl https://syntology.ai/api/ran/2003.03780.json

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

[![Syntology run record](https://syntology.ai/api/ran/2003.03780.svg)](https://syntology.ai/paper/2003.03780)

Connect an agent — have() is free.