Distributional Diffusion Models with Scoring Rules Article Swipe

PDF

Valentin De Bortoli , Alexandre Galashov , J. Swaroop Guntupalli , Guangyao Zhou , Kevin P. Murphy , Arthur Gretton , Arnaud Doucet ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2502.02483

Diffusion models generate high-quality synthetic data. They operate by defining a continuous-time forward process which gradually adds Gaussian noise to data until fully corrupted. The corresponding reverse process progressively "denoises" a Gaussian sample into a sample from the data distribution. However, generating high-quality outputs requires many discretization steps to obtain a faithful approximation of the reverse process. This is expensive and has motivated the development of many acceleration methods. We propose to accomplish sample generation by learning the posterior {\em distribution} of clean data samples given their noisy versions, instead of only the mean of this distribution. This allows us to sample from the probability transitions of the reverse process on a coarse time scale, significantly accelerating inference with minimal degradation of the quality of the output. This is accomplished by replacing the standard regression loss used to estimate conditional means with a scoring rule. We validate our method on image and robot trajectory generation, where we consistently outperform standard diffusion models at few discretization steps.

Related Topics

Computer Science

Economics

Physics

Concepts

Diffusion Econometrics Computer science Economics Thermodynamics Physics

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2502.02483
PDF: https://arxiv.org/pdf/2502.02483
OA Status: green
Cited By: 1
Related Works: 10
OpenAlex ID: https://openalex.org/W4407186801

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4407186801

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2502.02483

Digital Object Identifier
Title: Distributional Diffusion Models with Scoring Rules

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-02-04

Full publication date if available
Authors: Valentin De Bortoli, Alexandre Galashov, J. Swaroop Guntupalli, Guangyao Zhou, Kevin P. Murphy, Arthur Gretton, Arnaud Doucet

List of authors in order
Landing page: https://arxiv.org/abs/2502.02483

Publisher landing page
PDF URL: https://arxiv.org/pdf/2502.02483

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/2502.02483

Direct OA link when available
Concepts: Diffusion, Econometrics, Computer science, Economics, Thermodynamics, Physics

Top concepts (fields/topics) attached by OpenAlex
Cited by: 1

Total citation count in OpenAlex
Citations by year (recent): 2025: 1

Per-year citation counts (last 5 years)
Related works (count): 10

Other works algorithmically related by OpenAlex

Full payload

id	https://openalex.org/W4407186801
doi	https://doi.org/10.48550/arxiv.2502.02483
ids.doi	https://doi.org/10.48550/arxiv.2502.02483
ids.openalex	https://openalex.org/W4407186801
fwci
type	preprint
title	Distributional Diffusion Models with Scoring Rules
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
topics[0].id	https://openalex.org/T12011
topics[0].field.id	https://openalex.org/fields/33
topics[0].field.display_name	Social Sciences
topics[0].score	0.5336999893188477
topics[0].domain.id	https://openalex.org/domains/2
topics[0].domain.display_name	Social Sciences
topics[0].subfield.id	https://openalex.org/subfields/3317
topics[0].subfield.display_name	Demography
topics[0].display_name	Insurance, Mortality, Demography, Risk Management
is_xpac	False
apc_list
apc_paid
concepts[0].id	https://openalex.org/C69357855
concepts[0].level	2
concepts[0].score	0.5921759009361267
concepts[0].wikidata	https://www.wikidata.org/wiki/Q163214
concepts[0].display_name	Diffusion
concepts[1].id	https://openalex.org/C149782125
concepts[1].level	1
concepts[1].score	0.4213532507419586
concepts[1].wikidata	https://www.wikidata.org/wiki/Q160039
concepts[1].display_name	Econometrics
concepts[2].id	https://openalex.org/C41008148
concepts[2].level	0
concepts[2].score	0.4156843423843384
concepts[2].wikidata	https://www.wikidata.org/wiki/Q21198
concepts[2].display_name	Computer science
concepts[3].id	https://openalex.org/C162324750
concepts[3].level	0
concepts[3].score	0.2803223133087158
concepts[3].wikidata	https://www.wikidata.org/wiki/Q8134
concepts[3].display_name	Economics
concepts[4].id	https://openalex.org/C97355855
concepts[4].level	1
concepts[4].score	0.13330036401748657
concepts[4].wikidata	https://www.wikidata.org/wiki/Q11473
concepts[4].display_name	Thermodynamics
concepts[5].id	https://openalex.org/C121332964
concepts[5].level	0
concepts[5].score	0.0932302176952362
concepts[5].wikidata	https://www.wikidata.org/wiki/Q413
concepts[5].display_name	Physics
keywords[0].id	https://openalex.org/keywords/diffusion
keywords[0].score	0.5921759009361267
keywords[0].display_name	Diffusion
keywords[1].id	https://openalex.org/keywords/econometrics
keywords[1].score	0.4213532507419586
keywords[1].display_name	Econometrics
keywords[2].id	https://openalex.org/keywords/computer-science
keywords[2].score	0.4156843423843384
keywords[2].display_name	Computer science
keywords[3].id	https://openalex.org/keywords/economics
keywords[3].score	0.2803223133087158
keywords[3].display_name	Economics
keywords[4].id	https://openalex.org/keywords/thermodynamics
keywords[4].score	0.13330036401748657
keywords[4].display_name	Thermodynamics
keywords[5].id	https://openalex.org/keywords/physics
keywords[5].score	0.0932302176952362
keywords[5].display_name	Physics
language	en
locations[0].id	pmh:oai:arXiv.org:2502.02483
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/2502.02483
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/2502.02483
locations[1].id	doi:10.48550/arxiv.2502.02483
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	cc-by
locations[1].pdf_url
locations[1].version
locations[1].raw_type	article
locations[1].license_id	https://openalex.org/licenses/cc-by
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.2502.02483
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5050614717
authorships[0].author.orcid	https://orcid.org/0000-0002-7163-5391
authorships[0].author.display_name	Valentin De Bortoli
authorships[0].author_position	first
authorships[0].raw_author_name	De Bortoli, Valentin
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5067400056
authorships[1].author.orcid
authorships[1].author.display_name	Alexandre Galashov
authorships[1].author_position	middle
authorships[1].raw_author_name	Galashov, Alexandre
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5009263663
authorships[2].author.orcid	https://orcid.org/0000-0002-0677-5590
authorships[2].author.display_name	J. Swaroop Guntupalli
authorships[2].author_position	middle
authorships[2].raw_author_name	Guntupalli, J. Swaroop
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5076644778
authorships[3].author.orcid	https://orcid.org/0000-0003-0809-5799
authorships[3].author.display_name	Guangyao Zhou
authorships[3].author_position	middle
authorships[3].raw_author_name	Zhou, Guangyao
authorships[3].is_corresponding	False
authorships[4].author.id	https://openalex.org/A5032986126
authorships[4].author.orcid	https://orcid.org/0000-0002-8819-7834
authorships[4].author.display_name	Kevin P. Murphy
authorships[4].author_position	middle
authorships[4].raw_author_name	Murphy, Kevin
authorships[4].is_corresponding	False
authorships[5].author.id	https://openalex.org/A5004160420
authorships[5].author.orcid	https://orcid.org/0000-0003-3169-7624
authorships[5].author.display_name	Arthur Gretton
authorships[5].author_position	middle
authorships[5].raw_author_name	Gretton, Arthur
authorships[5].is_corresponding	False
authorships[6].author.id	https://openalex.org/A5091677854
authorships[6].author.orcid	https://orcid.org/0000-0002-7662-419X
authorships[6].author.display_name	Arnaud Doucet
authorships[6].author_position	last
authorships[6].raw_author_name	Doucet, Arnaud
authorships[6].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/2502.02483
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	Distributional Diffusion Models with Scoring Rules
has_fulltext	False
is_retracted	False
updated_date	2025-11-06T06:51:31.235846
primary_topic.id	https://openalex.org/T12011
primary_topic.field.id	https://openalex.org/fields/33
primary_topic.field.display_name	Social Sciences
primary_topic.score	0.5336999893188477
primary_topic.domain.id	https://openalex.org/domains/2
primary_topic.domain.display_name	Social Sciences
primary_topic.subfield.id	https://openalex.org/subfields/3317
primary_topic.subfield.display_name	Demography
primary_topic.display_name	Insurance, Mortality, Demography, Risk Management
related_works	https://openalex.org/W4391375266, https://openalex.org/W2899084033, https://openalex.org/W2748952813, https://openalex.org/W2390279801, https://openalex.org/W4391913857, https://openalex.org/W2358668433, https://openalex.org/W4396701345, https://openalex.org/W2376932109, https://openalex.org/W2001405890, https://openalex.org/W4396696052
cited_by_count	1
counts_by_year[0].year	2025
counts_by_year[0].cited_by_count	1
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2502.02483
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/2502.02483
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/2502.02483
primary_location.id	pmh:oai:arXiv.org:2502.02483
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/2502.02483
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/2502.02483
publication_date	2025-02-04
publication_year	2025
referenced_works_count	0
abstract_inverted_index.a	10, 30, 34, 50, 111, 142
abstract_inverted_index.We	69, 145
abstract_inverted_index.at	162
abstract_inverted_index.by	8, 75, 130
abstract_inverted_index.is	58, 128
abstract_inverted_index.of	53, 65, 81, 90, 94, 106, 121, 124
abstract_inverted_index.on	110, 149
abstract_inverted_index.to	19, 48, 71, 100, 137
abstract_inverted_index.us	99
abstract_inverted_index.we	156
abstract_inverted_index.The	24
abstract_inverted_index.and	60, 151
abstract_inverted_index.few	163
abstract_inverted_index.has	61
abstract_inverted_index.our	147
abstract_inverted_index.the	37, 54, 63, 77, 92, 103, 107, 122, 125, 132
abstract_inverted_index.They	6
abstract_inverted_index.This	57, 97, 127
abstract_inverted_index.adds	16
abstract_inverted_index.data	20, 38, 83
abstract_inverted_index.from	36, 102
abstract_inverted_index.into	33
abstract_inverted_index.loss	135
abstract_inverted_index.many	45, 66
abstract_inverted_index.mean	93
abstract_inverted_index.only	91
abstract_inverted_index.this	95
abstract_inverted_index.time	113
abstract_inverted_index.used	136
abstract_inverted_index.with	118, 141
abstract_inverted_index.{\em	79
abstract_inverted_index.clean	82
abstract_inverted_index.data.	5
abstract_inverted_index.fully	22
abstract_inverted_index.given	85
abstract_inverted_index.image	150
abstract_inverted_index.means	140
abstract_inverted_index.noise	18
abstract_inverted_index.noisy	87
abstract_inverted_index.robot	152
abstract_inverted_index.rule.	144
abstract_inverted_index.steps	47
abstract_inverted_index.their	86
abstract_inverted_index.until	21
abstract_inverted_index.where	155
abstract_inverted_index.which	14
abstract_inverted_index.allows	98
abstract_inverted_index.coarse	112
abstract_inverted_index.method	148
abstract_inverted_index.models	1, 161
abstract_inverted_index.obtain	49
abstract_inverted_index.sample	32, 35, 73, 101
abstract_inverted_index.scale,	114
abstract_inverted_index.steps.	165
abstract_inverted_index.forward	12
abstract_inverted_index.instead	89
abstract_inverted_index.minimal	119
abstract_inverted_index.operate	7
abstract_inverted_index.output.	126
abstract_inverted_index.outputs	43
abstract_inverted_index.process	13, 27, 109
abstract_inverted_index.propose	70
abstract_inverted_index.quality	123
abstract_inverted_index.reverse	26, 55, 108
abstract_inverted_index.samples	84
abstract_inverted_index.scoring	143
abstract_inverted_index.Gaussian	17, 31
abstract_inverted_index.However,	40
abstract_inverted_index.defining	9
abstract_inverted_index.estimate	138
abstract_inverted_index.faithful	51
abstract_inverted_index.generate	2
abstract_inverted_index.learning	76
abstract_inverted_index.methods.	68
abstract_inverted_index.process.	56
abstract_inverted_index.requires	44
abstract_inverted_index.standard	133, 159
abstract_inverted_index.validate	146
abstract_inverted_index.Diffusion	0
abstract_inverted_index.diffusion	160
abstract_inverted_index.expensive	59
abstract_inverted_index.gradually	15
abstract_inverted_index.inference	117
abstract_inverted_index.motivated	62
abstract_inverted_index.posterior	78
abstract_inverted_index.replacing	131
abstract_inverted_index.synthetic	4
abstract_inverted_index.versions,	88
abstract_inverted_index."denoises"	29
abstract_inverted_index.accomplish	72
abstract_inverted_index.corrupted.	23
abstract_inverted_index.generating	41
abstract_inverted_index.generation	74
abstract_inverted_index.outperform	158
abstract_inverted_index.regression	134
abstract_inverted_index.trajectory	153
abstract_inverted_index.conditional	139
abstract_inverted_index.degradation	120
abstract_inverted_index.development	64
abstract_inverted_index.generation,	154
abstract_inverted_index.probability	104
abstract_inverted_index.transitions	105
abstract_inverted_index.accelerating	116
abstract_inverted_index.acceleration	67
abstract_inverted_index.accomplished	129
abstract_inverted_index.consistently	157
abstract_inverted_index.high-quality	3, 42
abstract_inverted_index.approximation	52
abstract_inverted_index.corresponding	25
abstract_inverted_index.distribution.	39, 96
abstract_inverted_index.distribution}	80
abstract_inverted_index.progressively	28
abstract_inverted_index.significantly	115
abstract_inverted_index.discretization	46, 164
abstract_inverted_index.continuous-time	11
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	7
citation_normalized_percentile