The DOI matches its Crossref record; 61 of the 61 references it cites were found in a registry or an index.
Separating Style and Content with Bilinear Models
Abstract
Perceptual systems routinely separate “content” from “style,” classifying familiar words spoken in an unfamiliar accent, identifying a font or handwriting style across letters, or recognizing a familiar face or object seen under unfamiliar viewing conditions. Yet a general and tractable computational model of this ability to untangle the underlying factors of perceptual observations remains elusive (Hofstadter, 1985). Existing factor models (Mardia, Kent, & Bibby, 1979; Hinton & Zemel, 1994; Ghahramani, 1995; Bell & Sejnowski, 1995; Hinton, Dayan, Frey, & Neal, 1995; Dayan, Hinton, Neal, & Zemel, 1995; Hinton & Ghahramani, 1997) are either insufficiently rich to capture the complex interactions of perceptually meaningful factors such as phoneme and speaker accent or letter and font, or do not allow efficient learning algorithms. We present a general framework for learning to solve two-factor tasks using bilinear models, which provide sufficiently expressive representations of factor interactions but can nonetheless be fit to data using efficient algorithms based on the singular value decomposition and expectation-maximization. We report promising results on three different tasks in three different perceptual domains: spoken vowel classification with a benchmark multi-speaker database, extrapolation of fonts to unseen letters, and translation of faces to novel illuminants.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“GeAN also uses the bilinear model [29,30] to distil the knowledge interwoven in these representations by allowing each node and edge to collaboratively modulate each other's contributions through a shared learning space.”
From TVGeAN: Tensor Visibility Graph-Enhanced Attention Network for Versatile Multivariant Time Series Learning Tasks (2024) · cites this paper as [29]
Citations
Cites
Maximum Likelihood from Incomplete Data Via the EM Algorithm
Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977) · Journal of the Royal Statistical Society Series B: Statistical Methodology · doi:10.1111/j.2517-6161.1977.tb01600.x · Cited by 50,444
Neural networks for pattern recognition
Bishop, C. (1994) · Choice Reviews Online · doi:10.5860/choice.31-5500 · Cited by 18,537
Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. (1994) · Cited by 15,411
Turk, M., & Pentland, A. (1991) · Journal of Cognitive Neuroscience · doi:10.1162/jocn.1991.3.1.71 · Cited by 13,733
(1973) · CERN Document Server (European Organization for Nuclear Research) · Cited by 12,895
Numerical recipes in Pascal: the art of scientific computing
(1990) · Choice Reviews Online · doi:10.5860/choice.27-5158 · Cited by 12,040
An Information-Maximization Approach to Blind Separation and Blind Deconvolution
Bell, A. J., & Sejnowski, T. J. (1995) · Neural Computation · doi:10.1162/neco.1995.7.6.1129 · Cited by 8,986
Shape and motion from image streams under orthography: a factorization method
Tomasi, C., & Takeo Kanade, A. (1992) · International Journal of Computer Vision · doi:10.1007/bf00129684 · Cited by 2,879
Matrix Differential Calculus with Applications in Statistics and Econometrics.
Magnus, J. R., & Neudecker, H. (1988) · Biometrics · doi:10.2307/2531754 · Cited by 2,729
Classical Electrodynamics (2nd edn)
Brereton, M. (1976) · Physics Bulletin · doi:10.1088/0031-9112/27/8/041 · Cited by 2,612
Cited by
Sorted by citations
Items 801–839 of 839
Embedded Deep Bilinear Interactive Information and Selective Fusion for Multi-view LearningPreprint
Xu, J., Li, W., Shen, J., Liu, X., Zhou, P., Zhang, X., Yao, X., & Han, J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2007.06143 · Cited by 1
“Later, based on bilinear models [31] and general graph embedding framework [32], [33] introduces Generalized Multi-view Analysis (GMA).”
MethodSemi-Supervised Cross-Modal Retrieval Based on Discriminative Comapping
Liu, L., Dong, X. M., & Wang, T. (2020) · Complexity · doi:10.1155/2020/1462429
“In [20, 21], the bilinear model and generalized multiview analysis (GMA) have been proposed and performed well in the field of cross-modal retrieval.”
MethodLearning Image Manifolds from Local Features
Ma, Y., & Fu, Y. (2011) · CRC Press eBooks · doi:10.1201/b11431-16
Nuevos métodos para la combinación de características en procesamiento de imágenes
Carvalho, J. I. F. (2021) · doi:10.48035/tesis/2454/39683
Multilinear Dirichlet ProcessesPreprint
Li, X. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.08852
“As an extension of bilinear models [29], which was originally developed to untangle “content" and “style" factors in images, Multilinear factor analyzers (MLFA) [27] were developed to model data that is the result of multiplicative interactions of N groups of factors, with J n factors in each group.”
MethodAlpha-Integration Pooling for Convolutional Neural Networks
Eom, H., & Choi, H. (2021) · Journal of KIISE · doi:10.5626/jok.2021.48.7.774 · Cited by 1
“This pooling method has arithmetic average-pooling and bilinear pooling [20] as a special case, but max-pooling is not included.”
MethodVTuckeR: Multimodal Tucker Fusion for Scene Graph Generation
Jia, G., Yi, Y., & Kan, W. (2022) · doi:10.1145/3514105.3514111
“[15] This method can make full use of the interaction between vector elements.”
MethodLim, J., Frahm, J., & Monrose, F. (2022) · ACM Conference on Data and Application Security and Privacy (CODASPY) · doi:10.1145/3508398.3511498
“Tenenbaum and Freeman ([30] [31]) observe that by learning to factor observations of data into two independent factors of variation, style and content, models learn separate representations that can extrapolate style into novel content, classify content in different styles, and translate new content into new styles.”
BackgroundLow-Quality $$\textit{DanMu}$$ Detection via Eye-Tracking Patterns
Liu, X., He, W., Xu, T., et al. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-10989-8_20
Face and Facial Expression Recognition using Three Dimensional Data
Mpiperis, I., Malassiotis, S., & Strintzis, M. G. (2008) · Encyclopedia of Multimedia · doi:10.1007/978-0-387-78414-4_21
Multimodal Adversarial Learning
Osahor, U. (2022) · doi:10.33915/etd.11546
Multi-selection Attention for Multimodal Aspect-level Sentiment Classification
Miao, Y., Luo, R., Liu, T., Zhang, W., Cai, G., & Zhou, M. Q. (2022) · International Symposium on Parallel Architectures, Algorithms and Programming · doi:10.1109/paap56126.2022.10010454
“The bilinear pooling-based methods, also known as tensor fusion-based approaches, captures all interactions between modalities by the vector outer product [14].”
MethodTotal Disentanglement of Font Images into Style and Character Class FeaturesPreprint
Haraguchi, D., Shimoda, W., Yamaguchi, K., et al. (2022) · SSRN Electronic Journal · doi:10.2139/ssrn.4191248
“[34].”
Automatic Face Expression Neutralisation and Posture Correction via 3D Contrastive Learning
Wiem, G., & Douik, A. (2023) · Procedia Computer Science · doi:10.1016/j.procs.2023.10.011
Online video visual relation detection with hierarchical multi-modal fusion
He, Y., Gan, M.-G., & Ma, Q. (2024) · Multimedia Tools and Applications · doi:10.1007/s11042-023-15310-3
“The bilinear pooling method is also a commonly used fusion method, which performs the multiplicative interaction of all elements between the vectors by calculating the outer product of visual feature vector and language feature vector [12, 31, 39, 40, 42].”
Qi, J. (2024) · Applied and Computational Engineering · doi:10.54254/2755-2721/30/20230096
Day, S., Xiao, Z., Chaudhry, E., et al. (2024) · Axioms · doi:10.3390/axioms13030185
Neural Inertial Navigation Method for Wheeled Robots Based on Semi-supervised Co-training
Huang, F., Gao, M., Liu, Q., et al. (2024) · International Conference on Robotics, Intelligent Control and Artificial Intelligence (RICAI) · doi:10.1109/ricai64321.2024.10911499
“Due to the time-consuming iterative process of co-training and the need for a certain training speed, in the experimental stage, by comparing the convergence speeds and accuracies of four different network types, ResNet and Bi-LSTM network architectures are finally determined and selected respectively.”
InfluentialFast Normalization for Bilinear Pooling via Eigenvalue Regularization
Xu, S., Dong, H., Zhang, C., & Wang, C. (2025) · Applied Sciences · doi:10.3390/app15084155
Enhancing Domain Generalization for Robust Machine-Generated Text Detection
Park, S., Han, S., & Cha, M. (2025) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2025.3581694 · Cited by 1
“Representation disentanglement is the process of decomposing complex and entangled representations into independent features, each corresponding to a specific attribute [26].”
Style Matters: Intellectual Property Rights and the New Creative Economy
Isohanni, P., & Hietanen-Kunwald, P. (2025) · Perspectives in law, business and innovation · doi:10.1007/978-981-95-2058-9_10
Cloning with gesture expressivity
Rajagopal, M. K. (2012) · HAL · doi:10.70675/b1ce2501z6fcaz45ebza734z0c25881334c1
“To our knowledge, Tenanbaum and Freeman (2000) were the first who addressed style and content separation.”
BackgroundMethodFirst Place Solution to the MLCAS 2025 GWFSS Challenge: The Devil is in the Detail and Minority
Cao, S., Hu, T., & Lu, H. (2025) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw69036.2025.00738
“(2) computes the bilinear kernel [17].”
Upadhya, J., Poudel, K., & Ranganathan, J. (2025) · IEEE International Conference on Future Machine Learning and Data Science (FMLDS) · doi:10.1109/fmlds67896.2025.00128
“1) Bilinear Fusion:: The bilinear fusion mechanism models second-order interactions between image and text modalities through a learnable bilinear transformation [13], [28].”
Response selection for end-to-end retrieval-based dialogue systems
Boussaha, B. E. A. (2019) · theses.fr (ABES) · doi:10.70675/81045dfazc0f4z4eb2zaa8ezc9afd8f02bba
Learnable factored image representation for visual discovery
Dalens, T. (2019) · doi:10.70675/0c7a59f5zd8cbz44beza828z05a0d6ba9769
“While it is possible to apply the bilinear factor separation directly on the image pixels [Tenenbaum and Freeman, 2000] with some success on simple images, such as letters or digits, this approach does not work well for complex natural images.”
BackgroundMethodResultInfluentialToward interpretable machine learning, with applications to large-scale industrial systems data
Mita, G. (2021) · theses.fr (ABES) · doi:10.70675/acfd6da1z8046z49d6z8c4ez2da37571763c
Feature selection, sparse coding and normalization for material image Classification
Xu, S. (2021) · theses.fr (ABES) · doi:10.70675/f9a5c7a1zedd2z4c6aza032z9c1687f5063c
Realization of a diagnostic aid in orthodontics by deep learning
Doan, T. T. (2021) · theses.fr (ABES) · doi:10.70675/0dd414c4z04afz4f12zbe07z0e5c7fdebb94
Representations for generalization in Reinforcement Learning
Bertoin, D. (2023) · theses.fr (ABES) · doi:10.70675/793c524dzf379z40c7z8575zd28a2125db2a
Task-Relevant Representation Decoupling for Visual Reinforcement Learning Generalization
Wang, J., Lin, Y., Hu, X., Xu, Q., Wang, S., Chen, Z., & Lv, K. (2026) · ACM Transactions on Multimedia Computing Communications and Applications · doi:10.1145/3820367
“Although there are similar decoupling ideas [9, 26, 30, 33] in the field of Computer Vision, most of them are supervised and require labeled datasets or even paired datasets.”
Properties of the Hypothesis Space and their Effect on Machine Learning
Dulek, R. H. (2013) · Utrecht University Repository (Utrecht University)
Metadata
- Authors
- Tenenbaum, J. B., & Freeman, W. T.
- Affiliations
- Massachusetts Institute of Technology (US)
- Mitsubishi Electric Research Laboratories (United States) (US)
- Venue
- Neural Computation
- Publisher
- MIT Press
- Published
- 2000-06-01
- Volume
- 12
- Issue
- 6
- Pages
- 1247-1283
- ISSN
- 0899-7667 (ISSN-L)
- 1530-888X
- OpenAlex
- W2170653751
- Semantic Scholar
- 7e85f7d59e37972ec52cbabfef0512588d87f125
- Cited by
- 840
- References
- 61
- Citations by year
- 2026: 2 · 2025: 17 · 2024: 20 · 2023: 31 · 2022: 38 · 2021: 61 · 2020: 58 · 2019: 70 · 2018: 49 · 2017: 39 · 2016: 56 · 2015: 34 · 2014: 27 · 2013: 39 · 2012: 30
- FWCI
- 13.29
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- Neural Computation (publishedVersion)
- PubMed (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Neural Networks and Applications
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face and Expression Recognition
- Physical Sciences › Computer Science › Artificial Intelligence › Speech Recognition and Synthesis
- Keywords
- bilinear model, factor models, perceptual learning, speaker accent, singular value decomposition, expectation-maximization algorithm, factor interactions
- MeSH terms
- Algorithms, Face, Humans, Learning, Pattern Recognition, Visual, Speech Perception, Linear Models, Neural Networks, Computer
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
