Evaluating Probabilistic Inference in Deep Learning: Beyond Marginal Predictions. Article Swipe
A fundamental challenge for any intelligent system is prediction: given some inputs $X_1,..,X_\tau$ can you predict outcomes $Y_1,.., Y_\tau$. The KL divergence $\mathbf{d}_{\mathrm{KL}}$ provides a natural measure of prediction quality, but the majority of deep learning research looks only at the marginal predictions per input $X_t$. In this technical report we propose a scoring rule $\mathbf{d}_{\mathrm{KL}}^\tau$, parameterized by $\tau \in \mathcal{N}$ that evaluates the joint predictions at $\tau$ inputs simultaneously. We show that the commonly-used $\tau=1$ can be insufficient to drive good decisions in many settings of interest. We also show that, as $\tau$ grows, performing well according to $\mathbf{d}_{\mathrm{KL}}^\tau$ recovers universal guarantees for any possible decision. Finally, we provide problem-dependent guidance on the scale of $\tau$ for which our score provides sufficient guarantees for good performance.
Related Topics
Concepts
Metadata
- Type
- preprint
- Language
- en
- Landing Page
- https://arxiv.org/pdf/2107.09224v1
- OA Status
- green
- Cited By
- 1
- References
- 6
- Related Works
- 20
- OpenAlex ID
- https://openalex.org/W3185734706
All OpenAlex metadata
Raw OpenAlex JSON
- OpenAlex ID
-
https://openalex.org/W3185734706Canonical identifier for this work in OpenAlex
- Title
-
Evaluating Probabilistic Inference in Deep Learning: Beyond Marginal Predictions.Work title
- Type
-
preprintOpenAlex work type
- Language
-
enPrimary language
- Publication year
-
2021Year of publication
- Publication date
-
2021-07-20Full publication date if available
- Authors
-
Xiuyuan Lu, Ian Osband, Benjamin Van Roy, Zheng WenList of authors in order
- Landing page
-
https://arxiv.org/pdf/2107.09224v1Publisher landing page
- Open access
-
YesWhether a free full text is available
- OA status
-
greenOpen access status per OpenAlex
- OA URL
-
https://arxiv.org/pdf/2107.09224v1Direct OA link when available
- Concepts
-
Parameterized complexity, Divergence (linguistics), Inference, Probabilistic logic, Quality (philosophy), Computer science, Scale (ratio), Artificial intelligence, Physics, Mathematics, Algorithm, Philosophy, Quantum mechanics, LinguisticsTop concepts (fields/topics) attached by OpenAlex
- Cited by
-
1Total citation count in OpenAlex
- Citations by year (recent)
-
2021: 1Per-year citation counts (last 5 years)
- References (count)
-
6Number of works referenced by this work
- Related works (count)
-
20Other works algorithmically related by OpenAlex
Full payload
| id | https://openalex.org/W3185734706 |
|---|---|
| doi | |
| ids.mag | 3185734706 |
| ids.openalex | https://openalex.org/W3185734706 |
| fwci | |
| type | preprint |
| title | Evaluating Probabilistic Inference in Deep Learning: Beyond Marginal Predictions. |
| biblio.issue | |
| biblio.volume | |
| biblio.last_page | |
| biblio.first_page | |
| topics[0].id | https://openalex.org/T12535 |
| topics[0].field.id | https://openalex.org/fields/17 |
| topics[0].field.display_name | Computer Science |
| topics[0].score | 0.9995999932289124 |
| topics[0].domain.id | https://openalex.org/domains/3 |
| topics[0].domain.display_name | Physical Sciences |
| topics[0].subfield.id | https://openalex.org/subfields/1702 |
| topics[0].subfield.display_name | Artificial Intelligence |
| topics[0].display_name | Machine Learning and Data Classification |
| topics[1].id | https://openalex.org/T11689 |
| topics[1].field.id | https://openalex.org/fields/17 |
| topics[1].field.display_name | Computer Science |
| topics[1].score | 0.9993000030517578 |
| topics[1].domain.id | https://openalex.org/domains/3 |
| topics[1].domain.display_name | Physical Sciences |
| topics[1].subfield.id | https://openalex.org/subfields/1702 |
| topics[1].subfield.display_name | Artificial Intelligence |
| topics[1].display_name | Adversarial Robustness in Machine Learning |
| 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 | 0.9993000030517578 |
| 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/C165464430 |
| concepts[0].level | 2 |
| concepts[0].score | 0.7427500486373901 |
| concepts[0].wikidata | https://www.wikidata.org/wiki/Q1570441 |
| concepts[0].display_name | Parameterized complexity |
| concepts[1].id | https://openalex.org/C207390915 |
| concepts[1].level | 2 |
| concepts[1].score | 0.5320635437965393 |
| concepts[1].wikidata | https://www.wikidata.org/wiki/Q1230525 |
| concepts[1].display_name | Divergence (linguistics) |
| concepts[2].id | https://openalex.org/C2776214188 |
| concepts[2].level | 2 |
| concepts[2].score | 0.5193518400192261 |
| concepts[2].wikidata | https://www.wikidata.org/wiki/Q408386 |
| concepts[2].display_name | Inference |
| concepts[3].id | https://openalex.org/C49937458 |
| concepts[3].level | 2 |
| concepts[3].score | 0.5169705748558044 |
| concepts[3].wikidata | https://www.wikidata.org/wiki/Q2599292 |
| concepts[3].display_name | Probabilistic logic |
| concepts[4].id | https://openalex.org/C2779530757 |
| concepts[4].level | 2 |
| concepts[4].score | 0.46586254239082336 |
| concepts[4].wikidata | https://www.wikidata.org/wiki/Q1207505 |
| concepts[4].display_name | Quality (philosophy) |
| concepts[5].id | https://openalex.org/C41008148 |
| concepts[5].level | 0 |
| concepts[5].score | 0.4431125223636627 |
| concepts[5].wikidata | https://www.wikidata.org/wiki/Q21198 |
| concepts[5].display_name | Computer science |
| concepts[6].id | https://openalex.org/C2778755073 |
| concepts[6].level | 2 |
| concepts[6].score | 0.4346567392349243 |
| concepts[6].wikidata | https://www.wikidata.org/wiki/Q10858537 |
| concepts[6].display_name | Scale (ratio) |
| concepts[7].id | https://openalex.org/C154945302 |
| concepts[7].level | 1 |
| concepts[7].score | 0.4022054672241211 |
| concepts[7].wikidata | https://www.wikidata.org/wiki/Q11660 |
| concepts[7].display_name | Artificial intelligence |
| concepts[8].id | https://openalex.org/C121332964 |
| concepts[8].level | 0 |
| concepts[8].score | 0.3812611997127533 |
| concepts[8].wikidata | https://www.wikidata.org/wiki/Q413 |
| concepts[8].display_name | Physics |
| concepts[9].id | https://openalex.org/C33923547 |
| concepts[9].level | 0 |
| concepts[9].score | 0.3513147830963135 |
| concepts[9].wikidata | https://www.wikidata.org/wiki/Q395 |
| concepts[9].display_name | Mathematics |
| concepts[10].id | https://openalex.org/C11413529 |
| concepts[10].level | 1 |
| concepts[10].score | 0.2535405158996582 |
| concepts[10].wikidata | https://www.wikidata.org/wiki/Q8366 |
| concepts[10].display_name | Algorithm |
| concepts[11].id | https://openalex.org/C138885662 |
| concepts[11].level | 0 |
| concepts[11].score | 0.18689197301864624 |
| concepts[11].wikidata | https://www.wikidata.org/wiki/Q5891 |
| concepts[11].display_name | Philosophy |
| concepts[12].id | https://openalex.org/C62520636 |
| concepts[12].level | 1 |
| concepts[12].score | 0.09453025460243225 |
| concepts[12].wikidata | https://www.wikidata.org/wiki/Q944 |
| concepts[12].display_name | Quantum mechanics |
| concepts[13].id | https://openalex.org/C41895202 |
| concepts[13].level | 1 |
| concepts[13].score | 0.0 |
| concepts[13].wikidata | https://www.wikidata.org/wiki/Q8162 |
| concepts[13].display_name | Linguistics |
| keywords[0].id | https://openalex.org/keywords/parameterized-complexity |
| keywords[0].score | 0.7427500486373901 |
| keywords[0].display_name | Parameterized complexity |
| keywords[1].id | https://openalex.org/keywords/divergence |
| keywords[1].score | 0.5320635437965393 |
| keywords[1].display_name | Divergence (linguistics) |
| keywords[2].id | https://openalex.org/keywords/inference |
| keywords[2].score | 0.5193518400192261 |
| keywords[2].display_name | Inference |
| keywords[3].id | https://openalex.org/keywords/probabilistic-logic |
| keywords[3].score | 0.5169705748558044 |
| keywords[3].display_name | Probabilistic logic |
| keywords[4].id | https://openalex.org/keywords/quality |
| keywords[4].score | 0.46586254239082336 |
| keywords[4].display_name | Quality (philosophy) |
| keywords[5].id | https://openalex.org/keywords/computer-science |
| keywords[5].score | 0.4431125223636627 |
| keywords[5].display_name | Computer science |
| keywords[6].id | https://openalex.org/keywords/scale |
| keywords[6].score | 0.4346567392349243 |
| keywords[6].display_name | Scale (ratio) |
| keywords[7].id | https://openalex.org/keywords/artificial-intelligence |
| keywords[7].score | 0.4022054672241211 |
| keywords[7].display_name | Artificial intelligence |
| keywords[8].id | https://openalex.org/keywords/physics |
| keywords[8].score | 0.3812611997127533 |
| keywords[8].display_name | Physics |
| keywords[9].id | https://openalex.org/keywords/mathematics |
| keywords[9].score | 0.3513147830963135 |
| keywords[9].display_name | Mathematics |
| keywords[10].id | https://openalex.org/keywords/algorithm |
| keywords[10].score | 0.2535405158996582 |
| keywords[10].display_name | Algorithm |
| keywords[11].id | https://openalex.org/keywords/philosophy |
| keywords[11].score | 0.18689197301864624 |
| keywords[11].display_name | Philosophy |
| keywords[12].id | https://openalex.org/keywords/quantum-mechanics |
| keywords[12].score | 0.09453025460243225 |
| keywords[12].display_name | Quantum mechanics |
| language | en |
| locations[0].id | mag:3185734706 |
| 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 | |
| locations[0].version | submittedVersion |
| locations[0].raw_type | |
| locations[0].license_id | |
| locations[0].is_accepted | False |
| locations[0].is_published | False |
| locations[0].raw_source_name | arXiv (Cornell University) |
| locations[0].landing_page_url | https://arxiv.org/pdf/2107.09224v1 |
| authorships[0].author.id | https://openalex.org/A5027358754 |
| authorships[0].author.orcid | https://orcid.org/0000-0001-6376-8584 |
| authorships[0].author.display_name | Xiuyuan Lu |
| authorships[0].author_position | first |
| authorships[0].raw_author_name | Xiuyuan Lu |
| authorships[0].is_corresponding | False |
| authorships[1].author.id | https://openalex.org/A5015899120 |
| authorships[1].author.orcid | |
| authorships[1].author.display_name | Ian Osband |
| authorships[1].author_position | middle |
| authorships[1].raw_author_name | Ian Osband |
| authorships[1].is_corresponding | False |
| authorships[2].author.id | https://openalex.org/A5045543562 |
| authorships[2].author.orcid | https://orcid.org/0000-0001-8364-3746 |
| authorships[2].author.display_name | Benjamin Van Roy |
| authorships[2].author_position | middle |
| authorships[2].raw_author_name | Benjamin Van Roy |
| authorships[2].is_corresponding | False |
| authorships[3].author.id | https://openalex.org/A5067555437 |
| authorships[3].author.orcid | https://orcid.org/0000-0001-6987-6463 |
| authorships[3].author.display_name | Zheng Wen |
| authorships[3].author_position | last |
| authorships[3].raw_author_name | Zheng Wen |
| authorships[3].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/2107.09224v1 |
| open_access.oa_status | green |
| open_access.any_repository_has_fulltext | False |
| created_date | 2025-10-10T00:00:00 |
| display_name | Evaluating Probabilistic Inference in Deep Learning: Beyond Marginal Predictions. |
| has_fulltext | False |
| is_retracted | False |
| updated_date | 2025-10-10T17:16:08.811792 |
| primary_topic.id | https://openalex.org/T12535 |
| primary_topic.field.id | https://openalex.org/fields/17 |
| primary_topic.field.display_name | Computer Science |
| primary_topic.score | 0.9995999932289124 |
| primary_topic.domain.id | https://openalex.org/domains/3 |
| primary_topic.domain.display_name | Physical Sciences |
| primary_topic.subfield.id | https://openalex.org/subfields/1702 |
| primary_topic.subfield.display_name | Artificial Intelligence |
| primary_topic.display_name | Machine Learning and Data Classification |
| related_works | https://openalex.org/W878499472, https://openalex.org/W2159992248, https://openalex.org/W2150181672, https://openalex.org/W3081694630, https://openalex.org/W2107128331, https://openalex.org/W2123542217, https://openalex.org/W2953174487, https://openalex.org/W2963191005, https://openalex.org/W2026501144, https://openalex.org/W1492170170, https://openalex.org/W3012857547, https://openalex.org/W811226091, https://openalex.org/W2790324363, https://openalex.org/W2952352874, https://openalex.org/W3112176259, https://openalex.org/W1560621288, https://openalex.org/W3174145161, https://openalex.org/W2999107413, https://openalex.org/W3172178011, https://openalex.org/W2951516712 |
| cited_by_count | 1 |
| counts_by_year[0].year | 2021 |
| counts_by_year[0].cited_by_count | 1 |
| locations_count | 1 |
| best_oa_location.id | mag:3185734706 |
| 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 | |
| best_oa_location.version | submittedVersion |
| best_oa_location.raw_type | |
| best_oa_location.license_id | |
| best_oa_location.is_accepted | False |
| best_oa_location.is_published | False |
| best_oa_location.raw_source_name | arXiv (Cornell University) |
| best_oa_location.landing_page_url | https://arxiv.org/pdf/2107.09224v1 |
| primary_location.id | mag:3185734706 |
| 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 | |
| primary_location.version | submittedVersion |
| primary_location.raw_type | |
| primary_location.license_id | |
| primary_location.is_accepted | False |
| primary_location.is_published | False |
| primary_location.raw_source_name | arXiv (Cornell University) |
| primary_location.landing_page_url | https://arxiv.org/pdf/2107.09224v1 |
| publication_date | 2021-07-20 |
| publication_year | 2021 |
| referenced_works | https://openalex.org/W2919115771, https://openalex.org/W2600383743, https://openalex.org/W2025720061, https://openalex.org/W2039522160, https://openalex.org/W3134238439, https://openalex.org/W2962901934 |
| referenced_works_count | 6 |
| abstract_inverted_index.A | 0 |
| abstract_inverted_index.a | 24, 52 |
| abstract_inverted_index.In | 46 |
| abstract_inverted_index.KL | 20 |
| abstract_inverted_index.We | 70, 88 |
| abstract_inverted_index.as | 92 |
| abstract_inverted_index.at | 39, 66 |
| abstract_inverted_index.be | 77 |
| abstract_inverted_index.by | 57 |
| abstract_inverted_index.in | 83 |
| abstract_inverted_index.is | 7 |
| abstract_inverted_index.of | 27, 33, 86, 115 |
| abstract_inverted_index.on | 112 |
| abstract_inverted_index.to | 79, 98 |
| abstract_inverted_index.we | 50, 108 |
| abstract_inverted_index.The | 19 |
| abstract_inverted_index.\in | 59 |
| abstract_inverted_index.any | 4, 104 |
| abstract_inverted_index.but | 30 |
| abstract_inverted_index.can | 13, 76 |
| abstract_inverted_index.for | 3, 103, 117, 124 |
| abstract_inverted_index.our | 119 |
| abstract_inverted_index.per | 43 |
| abstract_inverted_index.the | 31, 40, 63, 73, 113 |
| abstract_inverted_index.you | 14 |
| abstract_inverted_index.also | 89 |
| abstract_inverted_index.deep | 34 |
| abstract_inverted_index.good | 81, 125 |
| abstract_inverted_index.many | 84 |
| abstract_inverted_index.only | 38 |
| abstract_inverted_index.rule | 54 |
| abstract_inverted_index.show | 71, 90 |
| abstract_inverted_index.some | 10 |
| abstract_inverted_index.that | 61, 72 |
| abstract_inverted_index.this | 47 |
| abstract_inverted_index.well | 96 |
| abstract_inverted_index.$\tau | 58 |
| abstract_inverted_index.drive | 80 |
| abstract_inverted_index.given | 9 |
| abstract_inverted_index.input | 44 |
| abstract_inverted_index.joint | 64 |
| abstract_inverted_index.looks | 37 |
| abstract_inverted_index.scale | 114 |
| abstract_inverted_index.score | 120 |
| abstract_inverted_index.that, | 91 |
| abstract_inverted_index.which | 118 |
| abstract_inverted_index.$X_t$. | 45 |
| abstract_inverted_index.$\tau$ | 67, 93, 116 |
| abstract_inverted_index.grows, | 94 |
| abstract_inverted_index.inputs | 11, 68 |
| abstract_inverted_index.report | 49 |
| abstract_inverted_index.system | 6 |
| abstract_inverted_index.measure | 26 |
| abstract_inverted_index.natural | 25 |
| abstract_inverted_index.predict | 15 |
| abstract_inverted_index.propose | 51 |
| abstract_inverted_index.provide | 109 |
| abstract_inverted_index.scoring | 53 |
| abstract_inverted_index.$Y_1,.., | 17 |
| abstract_inverted_index.$\tau=1$ | 75 |
| abstract_inverted_index.Finally, | 107 |
| abstract_inverted_index.Y_\tau$. | 18 |
| abstract_inverted_index.guidance | 111 |
| abstract_inverted_index.learning | 35 |
| abstract_inverted_index.majority | 32 |
| abstract_inverted_index.marginal | 41 |
| abstract_inverted_index.outcomes | 16 |
| abstract_inverted_index.possible | 105 |
| abstract_inverted_index.provides | 23, 121 |
| abstract_inverted_index.quality, | 29 |
| abstract_inverted_index.recovers | 100 |
| abstract_inverted_index.research | 36 |
| abstract_inverted_index.settings | 85 |
| abstract_inverted_index.according | 97 |
| abstract_inverted_index.challenge | 2 |
| abstract_inverted_index.decision. | 106 |
| abstract_inverted_index.decisions | 82 |
| abstract_inverted_index.evaluates | 62 |
| abstract_inverted_index.interest. | 87 |
| abstract_inverted_index.technical | 48 |
| abstract_inverted_index.universal | 101 |
| abstract_inverted_index.divergence | 21 |
| abstract_inverted_index.guarantees | 102, 123 |
| abstract_inverted_index.performing | 95 |
| abstract_inverted_index.prediction | 28 |
| abstract_inverted_index.sufficient | 122 |
| abstract_inverted_index.fundamental | 1 |
| abstract_inverted_index.intelligent | 5 |
| abstract_inverted_index.prediction: | 8 |
| abstract_inverted_index.predictions | 42, 65 |
| abstract_inverted_index.\mathcal{N}$ | 60 |
| abstract_inverted_index.insufficient | 78 |
| abstract_inverted_index.performance. | 126 |
| abstract_inverted_index.commonly-used | 74 |
| abstract_inverted_index.parameterized | 56 |
| abstract_inverted_index.$X_1,..,X_\tau$ | 12 |
| abstract_inverted_index.simultaneously. | 69 |
| abstract_inverted_index.problem-dependent | 110 |
| abstract_inverted_index.$\mathbf{d}_{\mathrm{KL}}$ | 22 |
| abstract_inverted_index.$\mathbf{d}_{\mathrm{KL}}^\tau$ | 99 |
| abstract_inverted_index.$\mathbf{d}_{\mathrm{KL}}^\tau$, | 55 |
| cited_by_percentile_year | |
| countries_distinct_count | 0 |
| institutions_distinct_count | 4 |
| sustainable_development_goals[0].id | https://metadata.un.org/sdg/16 |
| sustainable_development_goals[0].score | 0.6499999761581421 |
| sustainable_development_goals[0].display_name | Peace, Justice and strong institutions |
| citation_normalized_percentile |