Pith. sign in

REVIEW 3 major objections 2 minor 1 cited by

A Flexible Approach to Augmenting a Bayesian VAR with Nonlinear Factors

T0 review · 3 major / 2 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that a Bayesian VAR augmented with a small number of nonparametric regression-tree factors captures common nonlinearities parsimoniously, stays computationally feasible in high dimensions, and preserves structural shock id

desk verdict The submission is a file mix-up: abstract is a Bayesian VAR paper, body is an unrelated face-recognition paper, so the current file is unreviewable despite a plausible abstract. read the letter →

arxiv 2508.13972 v2 pith:FMOVTFVV submitted 2025-08-19 econ.EM stat.ME

Todd Clark , Florian Huber , Gary Koop This is my paper
classification econ.EMstat.ME MSC 62M1062F1562G0891B84
keywords BayesianVARnonlinearfactorsregressiontreesfunctionalpoolingnonparametrictimeseriesstructuralshockidentificationMarkovchainMonteCarlomacroeconomicforecasting
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that a Bayesian vector autoregression can be made nonlinear without sacrificing scalability: instead of letting every coefficient drift or adding many equation-specific nonlinear terms, it inserts a small number of latent nonlinear factors, each built from regression trees, that are shared across equations. The intended payoff is functional pooling—common nonlinear departures from linearity get modeled parsimoniously—while the nonparametric form reduces the risk of misspecifying the shape of the nonlinearity. The authors further argue that estimation by equation-by-equation MCMC remains simple even in very high-dimensional systems, avoiding the computational bottleneck of time-varying parameter VARs, and that standard identification of structural economic shocks in factor models can be carried over to the nonlinear setting. Artificial-data and macroeconomic exercises are presented as evidence of the model's forecasting and structural usefulness.

What carries the argument

The central object is the nonlinear factor: a small set of unobserved latent factors, each a regression-tree function of the data, that enter the VAR equations through factor loadings. Functional pooling is the mechanism that does the work—because each factor is shared across multiple equations, the model captures common nonlinearities with few parameters; the regression-tree form supplies flexibility; and the factor structure makes the likelihood factorize sufficiently that the MCMC sampler can be run equation by equation. This same factor structure is what allows structural shock identification, since identifying restrictions developed for linear factor models can be applied to the factors

What would settle it

Simulate a multivariate DGP where every equation has an independent, equation-specific nonlinear function and no common factor. If the proposed factor-augmented BVAR still forecasts well and posterior diagnostics show no neglected nonlinearity, the claim is supported; if it misses the idiosyncratic nonlinearities or produces distorted impulse responses, the functional-pooling premise fails. A simpler check: compare the model's out-of-sample forecasts against a version with equation-specific nonlinear terms on data known to be idiosyncratically nonlinear.

Watch

Extended reading notes

Core claim

The central claim is that a single Bayesian VAR augmented with a few nonparametric, regression-tree nonlinear factors is a flexible and practical workhorse for nonlinear macroeconometrics. The factors are the carriers of nonlinearity: because they are shared by many variables, the model achieves functional pooling—a small number of factors describe nonlinear behavior that is common across equations—and because they are regression trees, the form of the nonlinearity is learned rather than imposed. The paper maintains that this combination is parsimonious, robust to misspecification, computationally tractable through equation-by-equation MCMC even in high dimensions, and compatible with existi

Load-bearing premise

The load-bearing premise is that a small number of shared nonlinear factors can represent the nonlinear departures of all the variables in the system; if nonlinearities are mostly idiosyncratic to individual variables, the factor restriction misspecifies the data process and the parsimony, efficiency, and identification advantages weaken.

Editorial extensions

If this is right

  • Macroeconomic forecasting with nonlinearities becomes feasible in systems with many variables, where TVP-VARs are computationally prohibitive.
  • Structural impulse-response analysis under sign, zero, or other identifying restrictions can be conducted in nonlinear models using familiar factor-model techniques.
  • Borrowing strength across equations through shared factors should improve estimation of common nonlinear effects, especially in moderately sized samples.
  • The nonparametric tree form reduces the chance that the model imposes the wrong functional shape on the nonlinearity, relative to parametric nonlinear BVARs.
  • Equation-by-equation MCMC makes the approach easy to extend to larger cross-sections without bespoke sampling schemes.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial note: the full text supplied with this record describes an unrelated paper (on face-image quality assessment), so this extraction rests only on the title and abstract; the manuscript body could not be reviewed.
  • If the functional-pooling assumption is the binding constraint, a natural extension is to allow loadings to be estimated with shrinkage priors so the data decide which variables share a nonlinear factor and which do not.
  • The same construction could be tested against a model with equation-specific nonlinear terms via out-of-sample forecasting or marginal likelihood comparison; the gap would quantify how much pooling costs when nonlinearities are idiosyncratic.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 2 minor

Summary. The abstract announces a Bayesian VAR augmented with nonlinear factors modeled nonparametrically by regression trees. It claims four advantages: parsimonious modeling of common nonlinearities through functional pooling; reduced misspecification risk from the nonparametric treatment; tractable equation-by-equation MCMC even in very high-dimensional systems; and straightforward adaptation of structural shock identification from linear factor models. It promises illustrative artificial and macroeconomic exercises. The submitted full text, however, is a different paper: ViT-FIQA, a computer-vision paper on face image quality assessment using Vision Transformers, identified in its running header as arXiv:2508.13957v3 [cs.CV]. That full text contains no Bayesian VAR, no regression-tree factor specification, no estimation algorithm, no macroeconomic or artificial data exercises, and no code. The paper as submitted therefore consists of an abstract whose central claims cannot be checked against any supporting derivations or results.

Significance. If the claims in the abstract were correct, this would be a potentially useful contribution to empirical macroeconomics: a scalable nonlinear BVAR with factor structure, functional pooling, equation-by-equation MCMC, and transferable structural identification would address real bottlenecks in TVP-VAR estimation. The proposed idea is interesting and the identification transfer claim is worth pursuing. However, the manuscript provides none of the necessary evidence: no model definition, no prior specification, no sampler, no convergence diagnostics, no empirical exercises, and no reproducibility artifacts. The significance of the claimed contribution therefore cannot be assessed from the submitted material.

major comments (3)
  1. [Full text, running header and Sections 1–6] The full text supplied is not the paper described in the abstract. The running header identifies it as arXiv:2508.13957v3 [cs.CV] (ViT-FIQA), and Sections 1–6 concern face image quality assessment using Vision Transformers. There is no model definition for the nonlinear-factor BVAR, no regression-tree prior structure, no MCMC algorithm, no convergence diagnostics, and no artificial or macroeconomic exercise. Every load-bearing claim in the abstract—functional pooling, low misspecification risk, equation-by-equation MCMC tractability, and transferable shock identification—is therefore unsupported by any auditable manuscript content. This is not a local gap; the object of review is absent.
  2. [Abstract, first and third claimed advantages] The abstract asserts that 'a small number of nonlinear factors are used to model common nonlinearities across variables' and that 'Bayesian computation using MCMC is straightforward even in very high-dimensional models.' These are substantive technical claims. The first requires a specification of how the factor functions enter each equation, a statement of priors on the trees, and evidence that the factor representation is identifiable rather than merely a reparameterization of idiosyncratic nonlinearities. The second requires a concrete sampler with full conditional distributions, blocking, and evidence on mixing and scaling with dimension. None of this appears anywhere in the submitted manuscript; in isolation the abstract is an assertion, not a result.
  3. [Abstract, 'Exercises involving artificial and macroeconomic data'] The abstract promises artificial and macroeconomic exercises. The submitted full text contains only face-image quality benchmarks (Tables 1 and 2) and related vision experiments, with no economic data, no forecasting evaluation, and no structural analysis. This is direct evidence that the empirical support promised for the central claims is missing from the submitted version.
minor comments (2)
  1. [Full text, page headers] The running header on the supplied pages shows arXiv:2508.13957v3, not the arXiv ID of the submission. If this is a packaging error, the correct PDF must be provided; the abstract alone cannot be treated as a standalone paper.
  2. [Full text, References] The reference list is entirely from the face-recognition literature (e.g., [7], [31], [46]) and contains no citations relevant to Bayesian VARs, nonparametric factor models, or regression trees. No related econometric work is discussed.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation chain present; the abstract's claims are unsupported because the full text is a different paper, which is an integrity issue, not circularity.

full rationale

The abstract for arXiv:2508.13972 describes a Bayesian VAR augmented with nonparametric regression-tree nonlinear factors and claims functional pooling, low misspecification risk, tractable equation-by-equation MCMC, and adaptable structural shock identification. The supplied full text, however, is the computer-vision paper 'ViT-FIQA: Assessing Face Image Quality using Vision Transformers' (running header 'arXiv:2508.13957v3 [cs.CV] 21 Aug 2025'). There is no model definition, no sampling algorithm, no estimating equation, and no artificial or macroeconomic exercise corresponding to the abstract's claims. Circularity analysis asks whether a claimed prediction or derivation reduces to its own inputs by construction, e.g. by a parameter fitted to the target quantity or by a load-bearing self-citation. Here, there is no derivation chain at all: the abstract's advantages (functional pooling, parsimonious nonlinearity, MCMC tractability, identification transfer) are bare assertions with no equations, fits, or external benchmarks in the manuscript to audit. The mismatch between abstract and body is a serious completeness/integrity problem—the central claims are entirely unsupported by the submitted text—but it is not an instance of self-definition, fitted-input-called-prediction, self-citation load-bearing, or the other enumerated circularity patterns. Accordingly, the circularity score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

Abstract-level ledger. The paper introduces no new physical entity; its nonlinear latent factors are standard factor-model machinery, though they carry identifying assumptions. The load-bearing domain assumptions are functional pooling (a small set of shared nonlinear factors suffices), adequacy of regression-tree approximation, convergence of the equation-by-equation MCMC sampler in high dimensions, and transferability of linear identification schemes to nonlinear factors. Free parameters such as the number of factors and tree shrinkage hyperparameters are not disclosed in the abstract and would need to be audited in the full text.

free parameters (3)
  • Number of nonlinear factors
    Not stated in the abstract; the parsimony claim depends on this number being small, and any choice would trade off fit against the functional-pooling advantage.
  • Regression tree hyperparameters (number of trees, depth, shrinkage)
    Not stated in the abstract; these control how nonlinear the estimated factors can be and are typically set by prior or tuning in BART-type models.
  • BVAR shrinkage hyperparameters (Minnesota-type priors)
    Standard in the authors' BVAR program but not specified in the abstract; they affect whether the linear core is dominated by the nonlinear factors.
assumptions (4)
  • domain assumption A small number of common nonlinear factors captures the nonlinear departures in all VAR equations (functional pooling).
    This is the first claimed advantage and the load-bearing structural premise; if nonlinearities are idiosyncratic across variables, the factor model is misspecified.
  • domain assumption Regression trees can approximate the unknown nonlinear factor functions adequately under the chosen priors.
    Nonparametric approximation is claimed to reduce misspecification risk; this holds only if the tree ensemble is rich enough and the priors are suitable.
  • domain assumption Equation-by-equation MCMC converges and yields valid posteriors in high dimensions.
    The third advantage claims straightforward MCMC at scale; this presumes good mixing and valid conditional posteriors, which is a computational claim that must be checked in the full text.
  • domain assumption Structural identification schemes for linear factor models remain valid when factors are nonlinear.
    The fourth advantage asserts adaptation of identification from linear factor models; the abstract does not state what additional conditions the nonlinearity imposes.
invented entities (1)
  • Nonlinear latent factors (regression-tree functions)
    purpose: Model common nonlinearities across the VAR variables and enable functional pooling.
    They are latent constructs identified only through the model's priors and sampler; the abstract provides no external observable that isolates them. They are standard factor-model machinery, not a new physical entity, so the burden is modest.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Flexible Approach to Augmenting a Bayesian VAR with Nonlinear Factors." pith.science (2026). https://pith.science/paper/FMOVTFVV

@misc{pith2026250813972,
  author       = {Pith},
  title        = {Pith review of: A Flexible Approach to Augmenting a Bayesian VAR with Nonlinear Factors},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FMOVTFVV}},
  note         = {Machine review of arXiv:2508.13972}
}
read the original abstract

This paper proposes a vector autoregression augmented with nonlinear factors that are modeled nonparametrically using regression trees. There are four main advantages of our model. First, the use of factor methods ensures that departures from linearity are modeled parsimoniously. In particular, they exhibit functional pooling where a small number of nonlinear factors are used to model common nonlinearities across variables. Second, modeling potential nonlinearities nonparametrically lessens the risk of misspecification. Third, Bayesian computation using MCMC is straightforward even in very high-dimensional models, allowing for efficient, equation-by-equation estimation, thus avoiding computational bottlenecks that arise in popular alternatives such as the time-varying parameter VAR. Fourth, existing methods for identifying structural economic shocks in linear factor models can be adapted for the nonlinear case in a straightforward fashion using our model. Exercises involving artificial and macroeconomic data illustrate the properties of our model and its usefulness for forecasting and structural economic analysis.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. A Bayesian Gaussian Process Dynamic Factor Model

    econ.EM 2025-09 conditional novelty 6.0 of 10

    A Gaussian-process dynamic factor model with a linear VAR state equation produces modest out-of-sample forecast gains over linear DFMs and shows state-dependent global inflation dynamics.

Reference graph

Works this paper leans on

57 extracted references · 54 canonical work pages · cited by 1 Pith paper

  1. [1]

    Optimization-based improvement of face image quality assessment techniques

    ˇZiga Babnik, Naser Damer, and Vitomir ˇStruc. Optimization-based improvement of face image quality assessment techniques. In 11th International Workshop on Biometrics and F orensics, IWBF 2023, Barcelona, Spain, April 19-20, 2023, pages 1–6. IEEE, 2023. 1, 4

    work page 2023

  2. [2]

    Diffiqa: Face image quality assessment using denoising diffusion proba- bilistic models

    ˇZiga Babnik, Peter Peer, and Vitomir ˇStruc. Diffiqa: Face image quality assessment using denoising diffusion proba- bilistic models. In 2023 IEEE International Joint Conference on Biometrics (IJCB), pages 1–10, 2023. 2, 4, 5, 6

    work page 2023

  3. [3]

    eDifFIQA: To- wards Efficient Face Image Quality Assessment based on Denoising Diffusion Probabilistic Models

    ˇZiga Babnik, Peter Peer, and Vitomir ˇStruc. eDifFIQA: To- wards Efficient Face Image Quality Assessment based on Denoising Diffusion Probabilistic Models. IEEE Trans- actions on Biometrics, Behavior , and Identity Science (TBIOM), 2024. 2, 4, 5, 6

    work page 2024

  4. [4]

    Lacey Best-Rowden and Anil K. Jain. Learning face image quality from human assessments. IEEE Trans. Inf. F orensics Secur ., 13(12):3064–3077, 2018. 2

    work page 2018

  5. [5]

    Deep neural net- works for no-reference and full-reference image quality as- sessment

    Sebastian Bosse, Dominique Maniry, Klaus-Robert M ¨uller, Thomas Wiegand, and Wojciech Samek. Deep neural net- works for no-reference and full-reference image quality as- sessment. IEEE Trans. Image Process. , 27(1):206–219,

    work page

  6. [6]

    Elasticface: Elastic margin loss for deep face recognition

    Fadi Boutros, Naser Damer, Florian Kirchbuchner, and Ar- jan Kuijper. Elasticface: Elastic margin loss for deep face recognition. In IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPR Workshops 2022, New Orleans, LA, USA, June 19-20, 2022, pages 1577–1586. IEEE, 2022. 4

    work page 2022

  7. [7]

    CR-FIQA: face image quality assessment by learning sample relative classifiability

    Fadi Boutros, Meiling Fang, Marcel Klemt, Biying Fu, and Naser Damer. CR-FIQA: face image quality assessment by learning sample relative classifiability. In IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, CVPR 2023, V ancouver , BC, Canada, June 17-24, 2023 , pages 5836–5845. IEEE, 2023. 1, 2, 3, 4, 5, 6

    work page 2023

  8. [8]

    Face image quality assessment based on learning to rank

    Jiansheng Chen, Yu Deng, Gaocheng Bai, and Guangda Su. Face image quality assessment based on learning to rank. IEEE Signal Process. Lett., 22(1):90–94, 2015. 2, 5, 6

    work page 2015

Show all 57 references
  1. [9]

    Transface: Calibrating trans- former training for face recognition from a data-centric per- spective

    Jun Dan, Yang Liu, Haoyu Xie, Jiankang Deng, Haoran Xie, Xuansong Xie, and Baigui Sun. Transface: Calibrating trans- former training for face recognition from a data-centric per- spective. In ICCV, pages 20585–20596, 2023. 2, 3, 4

    2023

  2. [10]

    Arcface: Additive angular margin loss for deep face recognition

    Jiankang Deng, Jia Guo, Niannan Xue, and Stefanos Zafeiriou. Arcface: Additive angular margin loss for deep face recognition. In IEEE Conference on Computer Vi- sion and Pattern Recognition, CVPR 2019, Long Beach, CA, USA, June 16-20, 2019, pages 4690–4699. Computer Vision Fou...

    2019

  3. [11]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition ...

    2021

  4. [12]

    Age and gender estimation of unfiltered faces

    Eran Eidinger, Roee Enbar, and Tal Hassner. Age and gender estimation of unfiltered faces. IEEE Trans. Inf. F orensics Secur ., 9(12):2170–2179, 2014. 4, 5, 6

    2014

  5. [13]

    Best practice technical guidelines for automated border control (abc) systems, 2015

    Frontex. Best practice technical guidelines for automated border control (abc) systems, 2015. 4

    2015

  6. [14]

    A deep insight into measuring face image utility with general and face-specific image quality metrics

    Biying Fu, Cong Chen, Olaf Henniger, and Naser Damer. A deep insight into measuring face image utility with general and face-specific image quality metrics. In IEEE/CVF Win- ter Conference on Applications of Computer Vision, WACV 2022, Waikoloa, HI, USA, January 3-8, 2022 , pa...

    2022

  7. [15]

    Grother and E

    P. Grother and E. Tabassi. Performance of biometric quality measures. IEEE Trans. on Pattern Analysis and Machine Intelligence, 29(4):531–543, 2007. 4

    2007

  8. [16]

    Grother, M

    P. Grother, M. Ngan A. Hom, and K. Hanaoka. Ongoing face recognition vendor test (frvt) part 5: Face image quality assessment (4th draft). In National Institute of Standards and Technology. Tech. Rep., Sep. 2021. 1

    2021

  9. [17]

    Ms-celeb-1m: A dataset and benchmark for large-scale face recognition

    Yandong Guo, Lei Zhang, Yuxiao Hu, Xiaodong He, and Jianfeng Gao. Ms-celeb-1m: A dataset and benchmark for large-scale face recognition. In Computer Vision - ECCV 2016 - 14th European Conference, Amsterdam, The Nether- lands, October 11-14, 2016, Proceedings, Part III, pages 87–

    2016

  10. [18]

    Faceqnet: Qual- ity assessment for face recognition based on deep learning

    Javier Hernandez-Ortega, Javier Galbally, Julian Fi ´errez, Rudolf Haraksim, and Laurent Beslay. Faceqnet: Qual- ity assessment for face recognition based on deep learning. In 2019 International Conference on Biometrics, ICB 2019, Crete, Greece, June 4-7, 2019, pages 1–8. IEEE...

    2019

  11. [19]

    Biometric quality: Review and application to face recognition with faceqnet

    Javier Hernandez-Ortega, Javier Galbally, Julian Fi´errez, and Laurent Beslay. Biometric quality: Review and application to face recognition with faceqnet. CoRR, abs/2006.03298,

    2006 arXiv

  12. [20]

    Huang, Manu Ramesh, Tamara Berg, and Erik Learned-Miller

    Gary B. Huang, Manu Ramesh, Tamara Berg, and Erik Learned-Miller. Labeled faces in the wild: A database for studying face recognition in unconstrained environ- ments. Technical Report 07-49, University of Massachusetts, Amherst, 2007. 4, 5, 6

    2007

  13. [21]

    Curricularface: Adaptive curriculum learning loss for deep face recognition

    Yuge Huang, Yuhan Wang, Ying Tai, Xiaoming Liu, Pengcheng Shen, Shaoxin Li, Jilin Li, and Feiyue Huang. Curricularface: Adaptive curriculum learning loss for deep face recognition. In 2020 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, CVPR 2020, Seattle, WA...

    2020

  14. [22]

    ISO/IEC 29794-1:2016 In- formation technology - Biometric sample quality - Part 1: Framework

    ISO/IEC JTC1 SC37 Biometrics. ISO/IEC 29794-1:2016 In- formation technology - Biometric sample quality - Part 1: Framework. International Organization for Standardization,

    2016

  15. [23]

    ISO/IEC 19795-1:2021 In- formation technology — Biometric performance testing and reporting — Part 1: Principles and framework

    ISO/IEC JTC1 SC37 Biometrics. ISO/IEC 19795-1:2021 In- formation technology — Biometric performance testing and reporting — Part 1: Principles and framework. International Organization for Standardization, 2021. 4

    2021

  16. [24]

    Geunsu Kim, Gyudo Park, Soohyeok Kang, and Simon S. Woo. S-vit: Sparse vision transformer for accurate face recognition. In ACM SAC, pages 1130–1138, 2023. 2

    2023

  17. [25]

    Keypoint relative position encoding for face recog- nition

    Minchul Kim, Yiyang Su, Feng Liu, Anil Jain, and Xiaom- ing Liu. Keypoint relative position encoding for face recog- nition. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 244–255. IEEE, 2024. 2, 4

    2024

  18. [26]

    Berg, Wan-Yen Lo, Piotr Doll ´ar, and Ross B

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chlo´e Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C. Berg, Wan-Yen Lo, Piotr Doll ´ar, and Ross B. Girshick. Segment anything. In IEEE/CVF Inter- national Conference on Computer Vision, ICCV 202...

    2023

  19. [27]

    Cross-quality LFW: A database for analyzing cross- resolu- tion image face recognition in unconstrained environments

    Martin Knoche, Stefan H ¨ormann, and Gerhard Rigoll. Cross-quality LFW: A database for analyzing cross- resolu- tion image face recognition in unconstrained environments. In 16th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2021, Jodhpur , India,...

    2021

  20. [28]

    Grafiqs: Face image quality assessment using gradient magnitudes

    Jan Niklas Kolf, Naser Damer, and Fadi Boutros. Grafiqs: Face image quality assessment using gradient magnitudes. In 2024 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition Workshops (CVPRW) , pages 1490–1499,

    2024

  21. [29]

    Bagdanov

    Xialei Liu, Joost van de Weijer, and Andrew D. Bagdanov. Rankiqa: Learning from rankings for no-reference image quality assessment. In IEEE International Conference on Computer Vision, ICCV 2017, V enice, Italy, October 22-29, 2017, pages 1040–1049. IEEE Computer Society, 2017...

    2017

  22. [30]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019. OpenReview.net, 2019. 4

    2019

  23. [31]

    Magface: A universal representation for face recognition and quality assessment

    Qiang Meng, Shichao Zhao, Zhida Huang, and Feng Zhou. Magface: A universal representation for face recognition and quality assessment. In IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2021, virtual, June 19-25, 2021, pages 14225–14234. Computer Vision Founda...

    2021

  24. [32]

    No-reference image quality assessment in the spatial domain

    Anish Mittal, Anush Krishna Moorthy, and Alan Conrad Bovik. No-reference image quality assessment in the spatial domain. IEEE Trans. Image Process. , 21(12):4695–4708,

  25. [33]

    completely blind

    Anish Mittal, Rajiv Soundararajan, and Alan C Bovik. Mak- ing a “completely blind” image quality analyzer. IEEE SPL, 20(3):209–212, 2012. 1

    2012

  26. [34]

    Agedb: The first manually collected, in-the-wild age database

    Stylianos Moschoglou, Athanasios Papaioannou, Chris- tos Sagonas, Jiankang Deng, Irene Kotsia, and Stefanos Zafeiriou. Agedb: The first manually collected, in-the-wild age database. In 2017 IEEE CVPRW, CVPR Workshops 2017, Honolulu, HI, USA, July 21-26, 2017 , pages 1997–

    2017

  27. [35]

    Bui, and Khoa Luu

    Xuan-Bac Nguyen, Duc Toan Bui, Chi Nhan Duong, Tien D. Bui, and Khoa Luu. Clusformer: A transformer based clus- tering approach to unsupervised large-scale face and visual landmark recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognitio...

    2021

  28. [36]

    SDD-FIQA: unsupervised face image quality assess- ment with similarity distribution distance

    Fu-Zhao Ou, Xingyu Chen, Ruixin Zhang, Yuge Huang, Shaoxin Li, Jilin Li, Yong Li, Liujuan Cao, and Yuan-Gen Wang. SDD-FIQA: unsupervised face image quality assess- ment with similarity distribution distance. In IEEE Confer- ence on Computer Vision and Pattern Recognition, CVPR...

    2021

  29. [37]

    Clib-fiqa: Face image quality assessment with confidence calibration

    Fu-Zhao Ou, Chongyi Li, Shiqi Wang, and Sam Kwong. Clib-fiqa: Face image quality assessment with confidence calibration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 1694–1704, 2024. 4

    2024

  30. [38]

    Swinface: A multi-task trans- former for face recognition, expression recognition, age es- timation and attribute estimation

    Lixiong Qin, Mei Wang, Chao Deng, Ke Wang, Xi Chen, Jiani Hu, and Weihong Deng. Swinface: A multi-task trans- former for face recognition, expression recognition, age es- timation and attribute estimation. IEEE Trans. Circuits Syst. Video Technol., 34(4):2223–2234, 2024. 3, 4

    2024

  31. [39]

    Face image quality assessment: A literature survey

    Torsten Schlett, Christian Rathgeb, Olaf Henniger, Javier Galbally, Julian Fi ´errez, and Christoph Busch. Face image quality assessment: A literature survey. ACM Comput. Surv., 54(10s):210:1–210:49, 2022. 1

    2022

  32. [40]

    Tapia, and Christoph Busch

    Torsten Schlett, Christian Rathgeb, Juan E. Tapia, and Christoph Busch. Considerations on the evaluation of bio- metric quality assessment algorithms. IEEE Trans. Biom. Behav. Identity Sci., 6(1):54–67, 2024. 4

    2024

  33. [41]

    Patel, Rama Chellappa, and David W

    Soumyadip Sengupta, Jun-Cheng Chen, Carlos Domingo Castillo, Vishal M. Patel, Rama Chellappa, and David W. Ja- cobs. Frontal to profile face verification in the wild. In 2016 IEEE Winter Conference on Applications of Computer Vi- sion, WACV 2016, Lake Placid, NY, USA, March 7-...

    2016

  34. [42]

    Yichun Shi and Anil K. Jain. Probabilistic face embed- dings. In 2019 IEEE/CVF International Conference on Com- puter Vision, ICCV 2019, Seoul, Korea (South), October 27 - November 2, 2019 , pages 6901–6910. IEEE, 2019. 2, 5, 6

    2019

  35. [43]

    Part-based face recognition with vision transformers

    Zhonglin Sun and Georgios Tzimiropoulos. Part-based face recognition with vision transformers. In 33rd British Ma- chine Vision Conference 2022, BMVC 2022, London, UK, November 21-24, 2022, page 611. BMV A Press, 2022. 2

    2022

  36. [44]

    SER-FIQ: unsupervised estimation of face image quality based on stochastic embed- ding robustness

    Philipp Terh ¨orst, Jan Niklas Kolf, Naser Damer, Florian Kirchbuchner, and Arjan Kuijper. SER-FIQ: unsupervised estimation of face image quality based on stochastic embed- ding robustness. In 2020 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, CVPR 2020, Se...

    2020

  37. [45]

    Face quality estimation and its correlation to demographic and non-demographic bias in face recognition

    Philipp Terh ¨orst, Jan Niklas Kolf, Naser Damer, Florian Kirchbuchner, and Arjan Kuijper. Face quality estimation and its correlation to demographic and non-demographic bias in face recognition. In 2020 IEEE Int. Joint Conf. on Biometrics (IJCB), pages 1–11. IEEE, 2020. 1

    2020

  38. [46]

    Cosface: Large margin cosine loss for deep face recognition

    Hao Wang, Yitong Wang, Zheng Zhou, Xing Ji, Dihong Gong, Jingchao Zhou, Zhifeng Li, and Wei Liu. Cosface: Large margin cosine loss for deep face recognition. In 2018 IEEE Conference on Computer Vision and Pattern Recogni- tion, CVPR 2018, Salt Lake City, UT, USA, June 18-22, 2...

    2018

  39. [47]

    Face Analysis Technology Evaluation (FATE) Part 11: Face Image Quality Vector Assessment – Specific Image Defect Detection

    Joyce Yang, Patrick Grother, Mei Ngan, Kayee Hanaoka, and Austin Hom. Face Analysis Technology Evaluation (FATE) Part 11: Face Image Quality Vector Assessment – Specific Image Defect Detection. NIST Internal Report NIST IR 8485 DRAFT SUPPLEMENT, National Institute of Standards...

  40. [48]

    Lvface: Progressive cluster optimization for large vision models in face recogni- tion, 2025

    Jinghan You, Shanglin Li, Yuanrui Sun, Jiangchuan Wei, Mingyu Guo, Chao Feng, and Jiao Ran. Lvface: Progressive cluster optimization for large vision models in face recogni- tion, 2025. 3

    2025

  41. [49]

    Ni, and Heung-Yeung Shum

    Hao Zhang, Feng Li, Shilong Liu, Lei Zhang, Hang Su, Jun Zhu, Lionel M. Ni, and Heung-Yeung Shum. DINO: DETR with improved denoising anchor boxes for end-to-end ob- ject detection. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda,...

    2023

  42. [50]

    Joint face detection and alignment using multitask cascaded convolutional networks

    Kaipeng Zhang, Zhanpeng Zhang, Zhifeng Li, and Yu Qiao. Joint face detection and alignment using multitask cascaded convolutional networks. IEEE Signal Process. Lett. , 23(10): 1499–1503, 2016. 4

    2016

  43. [51]

    Zheng and W

    T. Zheng and W. Deng. Cross-pose lfw: A database for studying cross-pose face recognition in unconstrained en- vironments. Technical Report 18-01, Beijing University of Posts and Telecommunications, 2018. 4, 5, 6

    2018

  44. [52]

    Cross-age LFW: A database for studying cross-age face recognition in unconstrained environments

    Tianyue Zheng, Weihong Deng, and Jiani Hu. Cross-age LFW: A database for studying cross-age face recognition in unconstrained environments. CoRR, abs/1708.08197, 2017. 4, 5, 6

    2017 arXiv

  45. [53]

    Face transformer for recognition

    Yaoyao Zhong and Weihong Deng. Face transformer for recognition. arXiv preprint arXiv:2103.14803, 2021. 2

    2021 arXiv

  46. [54]

    FaceQAN: Face image quality assessment through adversarial noise explo- ration

    ˇZiga Babnik, Peter Peer, and Vitomir ˇStruc. FaceQAN: Face image quality assessment through adversarial noise explo- ration. In 26th International Conference on Pattern Recogni- tion (ICPR 2022) , pages 748–754, Montreal, Canada, 2022. IEEE. 2

    2022

  47. [2005]

    IEEE Computer Society, 2017. 4, 5, 6

    2017

  48. [2021]

    OpenReview.net, 2021. 2, 3, 4

    2021

  49. [2025]

    This publication is available free of charge

    Image Group, Information Access Division, Infor- mation Technology Laboratory. This publication is available free of charge. 2

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.