Explicit Regularisation in Gaussian Noise Injections Article Swipe

PDF

Alexander Camuto , Matthew Willetts , Umut Şimşekli , Stephen Roberts , Chris Holmes ·

YOU? · · 2020 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2007.07368

We study the regularisation induced in neural networks by Gaussian noise injections (GNIs). Though such injections have been extensively studied when applied to data, there have been few studies on understanding the regularising effect they induce when applied to network activations. Here we derive the explicit regulariser of GNIs, obtained by marginalising out the injected noise, and show that it penalises functions with high-frequency components in the Fourier domain; particularly in layers closer to a neural network's output. We show analytically and empirically that such regularisation produces calibrated classifiers with large classification margins.

Related Topics

Artificial Intelligence

Algorithm

Physics

Quantum Mechanics

Concepts

Noise (video) Gaussian noise Gaussian Mathematics Psychology Computer science Medicine Artificial intelligence Algorithm Physics Image (mathematics) Quantum mechanics

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2007.07368
PDF: https://arxiv.org/pdf/2007.07368
OA Status: green
Cited By: 35
References: 48
Related Works: 10
OpenAlex ID: https://openalex.org/W3042925324

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W3042925324

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2007.07368

Digital Object Identifier
Title: Explicit Regularisation in Gaussian Noise Injections

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2020

Year of publication
Publication date: 2020-07-14

Full publication date if available
Authors: Alexander Camuto, Matthew Willetts, Umut Şimşekli, Stephen Roberts, Chris Holmes

List of authors in order
Landing page: https://arxiv.org/abs/2007.07368

Publisher landing page
PDF URL: https://arxiv.org/pdf/2007.07368

Direct link to full text PDF
Open access: Yes

Whether a free full text is available
OA status: green

Open access status per OpenAlex
OA URL: https://arxiv.org/pdf/2007.07368

Direct OA link when available
Concepts: Noise (video), Gaussian noise, Gaussian, Mathematics, Psychology, Computer science, Medicine, Artificial intelligence, Algorithm, Physics, Image (mathematics), Quantum mechanics

Top concepts (fields/topics) attached by OpenAlex
Cited by: 35

Total citation count in OpenAlex
Citations by year (recent): 2025: 9, 2024: 7, 2023: 10, 2022: 4, 2021: 4

Per-year citation counts (last 5 years)
References (count): 48

Number of works referenced by this work
Related works (count): 10

Other works algorithmically related by OpenAlex

Full payload

id	https://openalex.org/W3042925324
doi	https://doi.org/10.48550/arxiv.2007.07368
ids.doi	https://doi.org/10.48550/arxiv.2007.07368
ids.mag	3042925324
ids.openalex	https://openalex.org/W3042925324
fwci	3.54773472
type	preprint
title	Explicit Regularisation in Gaussian Noise Injections
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
topics[0].id	https://openalex.org/T10500
topics[0].field.id	https://openalex.org/fields/22
topics[0].field.display_name	Engineering
topics[0].score	0.9991000294685364
topics[0].domain.id	https://openalex.org/domains/3
topics[0].domain.display_name	Physical Sciences
topics[0].subfield.id	https://openalex.org/subfields/2206
topics[0].subfield.display_name	Computational Mechanics
topics[0].display_name	Sparse and Compressive Sensing Techniques
topics[1].id	https://openalex.org/T10688
topics[1].field.id	https://openalex.org/fields/17
topics[1].field.display_name	Computer Science
topics[1].score	0.9983999729156494
topics[1].domain.id	https://openalex.org/domains/3
topics[1].domain.display_name	Physical Sciences
topics[1].subfield.id	https://openalex.org/subfields/1707
topics[1].subfield.display_name	Computer Vision and Pattern Recognition
topics[1].display_name	Image and Signal Denoising Methods
topics[2].id	https://openalex.org/T12814
topics[2].field.id	https://openalex.org/fields/17
topics[2].field.display_name	Computer Science
topics[2].score	0.996999979019165
topics[2].domain.id	https://openalex.org/domains/3
topics[2].domain.display_name	Physical Sciences
topics[2].subfield.id	https://openalex.org/subfields/1702
topics[2].subfield.display_name	Artificial Intelligence
topics[2].display_name	Gaussian Processes and Bayesian Inference
is_xpac	False
apc_list
apc_paid
concepts[0].id	https://openalex.org/C99498987
concepts[0].level	3
concepts[0].score	0.5535465478897095
concepts[0].wikidata	https://www.wikidata.org/wiki/Q2210247
concepts[0].display_name	Noise (video)
concepts[1].id	https://openalex.org/C4199805
concepts[1].level	2
concepts[1].score	0.4851142168045044
concepts[1].wikidata	https://www.wikidata.org/wiki/Q2725903
concepts[1].display_name	Gaussian noise
concepts[2].id	https://openalex.org/C163716315
concepts[2].level	2
concepts[2].score	0.43367812037467957
concepts[2].wikidata	https://www.wikidata.org/wiki/Q901177
concepts[2].display_name	Gaussian
concepts[3].id	https://openalex.org/C33923547
concepts[3].level	0
concepts[3].score	0.37364012002944946
concepts[3].wikidata	https://www.wikidata.org/wiki/Q395
concepts[3].display_name	Mathematics
concepts[4].id	https://openalex.org/C15744967
concepts[4].level	0
concepts[4].score	0.3528001606464386
concepts[4].wikidata	https://www.wikidata.org/wiki/Q9418
concepts[4].display_name	Psychology
concepts[5].id	https://openalex.org/C41008148
concepts[5].level	0
concepts[5].score	0.34416162967681885
concepts[5].wikidata	https://www.wikidata.org/wiki/Q21198
concepts[5].display_name	Computer science
concepts[6].id	https://openalex.org/C71924100
concepts[6].level	0
concepts[6].score	0.3255902826786041
concepts[6].wikidata	https://www.wikidata.org/wiki/Q11190
concepts[6].display_name	Medicine
concepts[7].id	https://openalex.org/C154945302
concepts[7].level	1
concepts[7].score	0.21080422401428223
concepts[7].wikidata	https://www.wikidata.org/wiki/Q11660
concepts[7].display_name	Artificial intelligence
concepts[8].id	https://openalex.org/C11413529
concepts[8].level	1
concepts[8].score	0.20586362481117249
concepts[8].wikidata	https://www.wikidata.org/wiki/Q8366
concepts[8].display_name	Algorithm
concepts[9].id	https://openalex.org/C121332964
concepts[9].level	0
concepts[9].score	0.18706974387168884
concepts[9].wikidata	https://www.wikidata.org/wiki/Q413
concepts[9].display_name	Physics
concepts[10].id	https://openalex.org/C115961682
concepts[10].level	2
concepts[10].score	0.0
concepts[10].wikidata	https://www.wikidata.org/wiki/Q860623
concepts[10].display_name	Image (mathematics)
concepts[11].id	https://openalex.org/C62520636
concepts[11].level	1
concepts[11].score	0.0
concepts[11].wikidata	https://www.wikidata.org/wiki/Q944
concepts[11].display_name	Quantum mechanics
keywords[0].id	https://openalex.org/keywords/noise
keywords[0].score	0.5535465478897095
keywords[0].display_name	Noise (video)
keywords[1].id	https://openalex.org/keywords/gaussian-noise
keywords[1].score	0.4851142168045044
keywords[1].display_name	Gaussian noise
keywords[2].id	https://openalex.org/keywords/gaussian
keywords[2].score	0.43367812037467957
keywords[2].display_name	Gaussian
keywords[3].id	https://openalex.org/keywords/mathematics
keywords[3].score	0.37364012002944946
keywords[3].display_name	Mathematics
keywords[4].id	https://openalex.org/keywords/psychology
keywords[4].score	0.3528001606464386
keywords[4].display_name	Psychology
keywords[5].id	https://openalex.org/keywords/computer-science
keywords[5].score	0.34416162967681885
keywords[5].display_name	Computer science
keywords[6].id	https://openalex.org/keywords/medicine
keywords[6].score	0.3255902826786041
keywords[6].display_name	Medicine
keywords[7].id	https://openalex.org/keywords/artificial-intelligence
keywords[7].score	0.21080422401428223
keywords[7].display_name	Artificial intelligence
keywords[8].id	https://openalex.org/keywords/algorithm
keywords[8].score	0.20586362481117249
keywords[8].display_name	Algorithm
keywords[9].id	https://openalex.org/keywords/physics
keywords[9].score	0.18706974387168884
keywords[9].display_name	Physics
language	en
locations[0].id	pmh:oai:arXiv.org:2007.07368
locations[0].is_oa	True
locations[0].source.id	https://openalex.org/S4306400194
locations[0].source.issn
locations[0].source.type	repository
locations[0].source.is_oa	True
locations[0].source.issn_l
locations[0].source.is_core	False
locations[0].source.is_in_doaj	False
locations[0].source.display_name	arXiv (Cornell University)
locations[0].source.host_organization	https://openalex.org/I205783295
locations[0].source.host_organization_name	Cornell University
locations[0].source.host_organization_lineage	https://openalex.org/I205783295
locations[0].license
locations[0].pdf_url	https://arxiv.org/pdf/2007.07368
locations[0].version	submittedVersion
locations[0].raw_type	text
locations[0].license_id
locations[0].is_accepted	False
locations[0].is_published	False
locations[0].raw_source_name
locations[0].landing_page_url	http://arxiv.org/abs/2007.07368
locations[1].id	doi:10.48550/arxiv.2007.07368
locations[1].is_oa	True
locations[1].source.id	https://openalex.org/S4306400194
locations[1].source.issn
locations[1].source.type	repository
locations[1].source.is_oa	True
locations[1].source.issn_l
locations[1].source.is_core	False
locations[1].source.is_in_doaj	False
locations[1].source.display_name	arXiv (Cornell University)
locations[1].source.host_organization	https://openalex.org/I205783295
locations[1].source.host_organization_name	Cornell University
locations[1].source.host_organization_lineage	https://openalex.org/I205783295
locations[1].license
locations[1].pdf_url
locations[1].version
locations[1].raw_type	article-journal
locations[1].license_id
locations[1].is_accepted	False
locations[1].is_published
locations[1].raw_source_name
locations[1].landing_page_url	https://doi.org/10.48550/arxiv.2007.07368
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5075596560
authorships[0].author.orcid
authorships[0].author.display_name	Alexander Camuto
authorships[0].author_position	first
authorships[0].raw_author_name	Alexander Camuto
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5019067441
authorships[1].author.orcid	https://orcid.org/0000-0001-6585-3950
authorships[1].author.display_name	Matthew Willetts
authorships[1].author_position	middle
authorships[1].raw_author_name	Matthew Willetts
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5053326908
authorships[2].author.orcid
authorships[2].author.display_name	Umut Şimşekli
authorships[2].countries	BE
authorships[2].affiliations[0].institution_ids	https://openalex.org/I2802330208
authorships[2].affiliations[0].raw_affiliation_string	Signal, Statistique et Apprentissage
authorships[2].affiliations[1].raw_affiliation_string	Institut Polytechnique de Paris
authorships[2].affiliations[2].raw_affiliation_string	Département Images, Données, Signal
authorships[2].institutions[0].id	https://openalex.org/I2802330208
authorships[2].institutions[0].ror	https://ror.org/0246a9012
authorships[2].institutions[0].type	government
authorships[2].institutions[0].lineage	https://openalex.org/I2802330208, https://openalex.org/I4210095879
authorships[2].institutions[0].country_code	BE
authorships[2].institutions[0].display_name	Statistics Belgium
authorships[2].author_position	middle
authorships[2].raw_author_name	Umut Şimşekli
authorships[2].is_corresponding	False
authorships[2].raw_affiliation_strings	Département Images, Données, Signal, Institut Polytechnique de Paris, Signal, Statistique et Apprentissage
authorships[3].author.id	https://openalex.org/A5044000346
authorships[3].author.orcid	https://orcid.org/0000-0002-6730-3108
authorships[3].author.display_name	Stephen Roberts
authorships[3].author_position	middle
authorships[3].raw_author_name	Stephen Roberts
authorships[3].is_corresponding	False
authorships[4].author.id	https://openalex.org/A5090278748
authorships[4].author.orcid	https://orcid.org/0000-0003-4532-3009
authorships[4].author.display_name	Chris Holmes
authorships[4].author_position	last
authorships[4].raw_author_name	Chris Holmes
authorships[4].is_corresponding	False
has_content.pdf	False
has_content.grobid_xml	False
is_paratext	False
open_access.is_oa	True
open_access.oa_url	https://arxiv.org/pdf/2007.07368
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	Explicit Regularisation in Gaussian Noise Injections
has_fulltext	False
is_retracted	False
updated_date	2025-11-06T06:51:31.235846
primary_topic.id	https://openalex.org/T10500
primary_topic.field.id	https://openalex.org/fields/22
primary_topic.field.display_name	Engineering
primary_topic.score	0.9991000294685364
primary_topic.domain.id	https://openalex.org/domains/3
primary_topic.domain.display_name	Physical Sciences
primary_topic.subfield.id	https://openalex.org/subfields/2206
primary_topic.subfield.display_name	Computational Mechanics
primary_topic.display_name	Sparse and Compressive Sensing Techniques
related_works	https://openalex.org/W2748952813, https://openalex.org/W4391375266, https://openalex.org/W1979597421, https://openalex.org/W2007980826, https://openalex.org/W2061531152, https://openalex.org/W3002753104, https://openalex.org/W1970319972, https://openalex.org/W2572355887, https://openalex.org/W2953254336, https://openalex.org/W2112030392
cited_by_count	35
counts_by_year[0].year	2025
counts_by_year[0].cited_by_count	9
counts_by_year[1].year	2024
counts_by_year[1].cited_by_count	7
counts_by_year[2].year	2023
counts_by_year[2].cited_by_count	10
counts_by_year[3].year	2022
counts_by_year[3].cited_by_count	4
counts_by_year[4].year	2021
counts_by_year[4].cited_by_count	4
counts_by_year[5].year	2020
counts_by_year[5].cited_by_count	1
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2007.07368
best_oa_location.is_oa	True
best_oa_location.source.id	https://openalex.org/S4306400194
best_oa_location.source.issn
best_oa_location.source.type	repository
best_oa_location.source.is_oa	True
best_oa_location.source.issn_l
best_oa_location.source.is_core	False
best_oa_location.source.is_in_doaj	False
best_oa_location.source.display_name	arXiv (Cornell University)
best_oa_location.source.host_organization	https://openalex.org/I205783295
best_oa_location.source.host_organization_name	Cornell University
best_oa_location.source.host_organization_lineage	https://openalex.org/I205783295
best_oa_location.license
best_oa_location.pdf_url	https://arxiv.org/pdf/2007.07368
best_oa_location.version	submittedVersion
best_oa_location.raw_type	text
best_oa_location.license_id
best_oa_location.is_accepted	False
best_oa_location.is_published	False
best_oa_location.raw_source_name
best_oa_location.landing_page_url	http://arxiv.org/abs/2007.07368
primary_location.id	pmh:oai:arXiv.org:2007.07368
primary_location.is_oa	True
primary_location.source.id	https://openalex.org/S4306400194
primary_location.source.issn
primary_location.source.type	repository
primary_location.source.is_oa	True
primary_location.source.issn_l
primary_location.source.is_core	False
primary_location.source.is_in_doaj	False
primary_location.source.display_name	arXiv (Cornell University)
primary_location.source.host_organization	https://openalex.org/I205783295
primary_location.source.host_organization_name	Cornell University
primary_location.source.host_organization_lineage	https://openalex.org/I205783295
primary_location.license
primary_location.pdf_url	https://arxiv.org/pdf/2007.07368
primary_location.version	submittedVersion
primary_location.raw_type	text
primary_location.license_id
primary_location.is_accepted	False
primary_location.is_published	False
primary_location.raw_source_name
primary_location.landing_page_url	http://arxiv.org/abs/2007.07368
publication_date	2020-07-14
publication_year	2020
referenced_works	https://openalex.org/W2962857907, https://openalex.org/W2962702169, https://openalex.org/W2753482910, https://openalex.org/W1519506695, https://openalex.org/W2963724148, https://openalex.org/W2626325961, https://openalex.org/W2152722485, https://openalex.org/W2897097528, https://openalex.org/W2071935776, https://openalex.org/W3007325053, https://openalex.org/W2963959597, https://openalex.org/W2963545688, https://openalex.org/W2962734330, https://openalex.org/W3092390915, https://openalex.org/W2964153729, https://openalex.org/W1988115241, https://openalex.org/W2146596719, https://openalex.org/W2047278710, https://openalex.org/W2254249950, https://openalex.org/W2023943903, https://openalex.org/W2098824882, https://openalex.org/W2111406701, https://openalex.org/W2800112096, https://openalex.org/W2971043187, https://openalex.org/W2809090039, https://openalex.org/W2964212410, https://openalex.org/W3137695714, https://openalex.org/W2149846618, https://openalex.org/W2529714286, https://openalex.org/W3105541982, https://openalex.org/W2952204734, https://openalex.org/W2768267830, https://openalex.org/W2605372163, https://openalex.org/W2911634294, https://openalex.org/W2044618041, https://openalex.org/W2095705004, https://openalex.org/W2996326647, https://openalex.org/W2963207607, https://openalex.org/W2962821226, https://openalex.org/W2964088238, https://openalex.org/W2891262038, https://openalex.org/W2962781217, https://openalex.org/W2963739978, https://openalex.org/W2996271040, https://openalex.org/W1862031973, https://openalex.org/W1826234144, https://openalex.org/W1599263113, https://openalex.org/W2243397390
referenced_works_count	48
abstract_inverted_index.a	74
abstract_inverted_index.We	0, 78
abstract_inverted_index.by	8, 50
abstract_inverted_index.in	5, 65, 70
abstract_inverted_index.it	59
abstract_inverted_index.of	47
abstract_inverted_index.on	29
abstract_inverted_index.to	22, 38, 73
abstract_inverted_index.we	42
abstract_inverted_index.and	56, 81
abstract_inverted_index.few	27
abstract_inverted_index.out	52
abstract_inverted_index.the	2, 31, 44, 53, 66
abstract_inverted_index.Here	41
abstract_inverted_index.been	17, 26
abstract_inverted_index.have	16, 25
abstract_inverted_index.show	57, 79
abstract_inverted_index.such	14, 84
abstract_inverted_index.that	58, 83
abstract_inverted_index.they	34
abstract_inverted_index.when	20, 36
abstract_inverted_index.with	62, 89
abstract_inverted_index.GNIs,	48
abstract_inverted_index.data,	23
abstract_inverted_index.large	90
abstract_inverted_index.noise	10
abstract_inverted_index.study	1
abstract_inverted_index.there	24
abstract_inverted_index.Though	13
abstract_inverted_index.closer	72
abstract_inverted_index.derive	43
abstract_inverted_index.effect	33
abstract_inverted_index.induce	35
abstract_inverted_index.layers	71
abstract_inverted_index.neural	6, 75
abstract_inverted_index.noise,	55
abstract_inverted_index.(GNIs).	12
abstract_inverted_index.Fourier	67
abstract_inverted_index.applied	21, 37
abstract_inverted_index.domain;	68
abstract_inverted_index.induced	4
abstract_inverted_index.network	39
abstract_inverted_index.output.	77
abstract_inverted_index.studied	19
abstract_inverted_index.studies	28
abstract_inverted_index.Gaussian	9
abstract_inverted_index.explicit	45
abstract_inverted_index.injected	54
abstract_inverted_index.margins.	92
abstract_inverted_index.networks	7
abstract_inverted_index.obtained	49
abstract_inverted_index.produces	86
abstract_inverted_index.functions	61
abstract_inverted_index.network's	76
abstract_inverted_index.penalises	60
abstract_inverted_index.calibrated	87
abstract_inverted_index.components	64
abstract_inverted_index.injections	11, 15
abstract_inverted_index.classifiers	88
abstract_inverted_index.empirically	82
abstract_inverted_index.extensively	18
abstract_inverted_index.regulariser	46
abstract_inverted_index.activations.	40
abstract_inverted_index.analytically	80
abstract_inverted_index.particularly	69
abstract_inverted_index.regularising	32
abstract_inverted_index.marginalising	51
abstract_inverted_index.understanding	30
abstract_inverted_index.classification	91
abstract_inverted_index.high-frequency	63
abstract_inverted_index.regularisation	3, 85
cited_by_percentile_year.max	99
cited_by_percentile_year.min	89
countries_distinct_count	1
institutions_distinct_count	5
sustainable_development_goals[0].id	https://metadata.un.org/sdg/10
sustainable_development_goals[0].score	0.7799999713897705
sustainable_development_goals[0].display_name	Reduced inequalities
citation_normalized_percentile.value	0.93371903
citation_normalized_percentile.is_in_top_1_percent	False
citation_normalized_percentile.is_in_top_10_percent	True