Pith. sign in

REVIEW 4 major objections 4 minor 43 references

Null Space Analysis for Class-Specific Discriminant Learning

T0 review · 4 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Null-space directions of the scatter matrices carry the discriminative power in class-specific learning, and new algorithms built on this identity outperform the standard CSDA baseline.

desk verdict Section IV's key inference is false (Nt=∅ doesn't imply Np=N⊥n), so the theory collapses, but the empirical CSDA variants and the step-by-step analysis are solid enough to warrant a careful referee. read the letter →

arxiv 1908.04562 v1 pith:7KBNJZBO submitted 2019-08-13 cs.LG stat.ML

Jenni Raitoharju , Alexandros Iosifidis This is my paper · ORCID
classification cs.LGstat.ML MSC 62H3068T1015A18
keywords Class-SpecificDiscriminantAnalysisNullspaceDimensionalityreductionSmallsamplesizeproblemMulti-modaldatadistributionsKernelmethodsFaceverification
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 claims that class-specific discriminant analysis (CSDA), which seeks directions that keep samples of the class of interest close to its mean while pushing everything else away, is best understood through the null spaces of the scatter matrices involved. In the small-sample-size regime, where the data dimension is at least as large as the number of training samples, the authors establish an identity between the null space of the total scatter matrix and the intersection of the null spaces of the intra-class and out-of-class scatter matrices. This identity justifies a three-step algorithm: remove the null space of the total scatter, compute the null space of the intra-class scatter, then maximize the out-of-class scatter within it. The resulting methods, called NCSDA and its orthogonal and regularized relatives, are reported to outperform the standard CSDA baseline on face and scene datasets and to remain competitive with more recent class-specific discriminant methods. The significance is that the small-sample-size problem, usually treated as a numerical obstacle, is recast as the source of the most informative discriminant directions.

What carries the argument

The central object is the null-space identity $\mathcal{N}_t = \mathcal{N}_p \cap \mathcal{N}_n$, where $\mathcal{N}_x$ denotes the set of directions annihilated by the symmetric scatter matrix $S_x$ (the null space). The identity is what licenses the recipe used throughout the paper: map the data to the row space of $S_t$ so that its null space disappears, then compute the null space of the intra-class scatter $S_p$, and finally maximize the out-of-class scatter $S_n$ inside that null space. The algorithmic machinery also includes several generalized eigenproblems for carrying out the second step in a numerically stable way, a whitening transformation that makes $S_t$ well-conditioned and so repairs the numerical gap between $\mathcal{N}_p$ and $\mathcal{N}_n^\perp$, and the cluster-based decomposition $S_n = S_{nw} + S_{nb}$ that underlies the heterogeneous variants.

What would settle it

Take $S_p = c(1,0,-2)(1,0,-2)^\top$ and $S_n = \mathrm{diag}(1,1,0)$ in $\mathbb{R}^3$ with $c>0$. Then $S_t = S_p+S_n$ has full rank, so $\mathcal{N}_t=\emptyset$, but $\mathcal{N}_p = \mathrm{span}\{e_2,(2,0,1)\}$ while $\mathcal{N}_n^\perp = \mathrm{span}\{e_1,e_2\}$; this concrete calculation would disprove the claimed consequence that a full-rank $S_t$ forces $\mathcal{N}_p=\mathcal{N}_n^\perp$.

Watch

Extended reading notes

Core claim

The central claim is that, after centering the data at the positive-class mean, the null space of the total scatter matrix $S_t = S_p + S_n$ equals the intersection of the null spaces of the intra-class scatter $S_p$ and the out-of-class scatter $S_n$, i.e., $\mathcal{N}_t = \mathcal{N}_p \cap \mathcal{N}_n$. When $S_t$ is full rank, this identity is said to force $\mathcal{N}_p = \mathcal{N}_n^\perp$, meaning that the directions which keep the positive class together are exactly the directions that spread the negative class out, so maximizing $S_n$ in the null space of $S_p$ captures all available discriminant information. The paper builds a family of algorithms on this identity: project onto the row space of $S_t$ to empty its null space, extract the null space of the projected $S_p$, and then rank the resulting directions by the out-of-class scatter they carry. A whitening step is shown to remove the numerical misalignment between the computed null space of $S_p$ and the row space of $S_n$, while its regularized version restores a meaningful ranking for low-dimensional projections. The heterogeneous extensions $S_n = S_{nw} + S_{nb}$ split the negative class into clusters, so that the algorithm pushes whole clusters away from the positive mean, with the number of clusters acting as a tunable parameter.

Load-bearing premise

The entire argument rests on the training-set size being no larger than the data dimensionality after preprocessing, so that the scatter matrices are singular and have non-trivial null spaces; outside this small-sample-size regime the claimed null-space alignment is not established.

Editorial extensions

If this is right

  • If the identity holds, the proposed NCSDA captures all discriminant directions that the standard CSDA criterion can express, with innate subspace dimensionality equal to the rank of the out-of-class scatter matrix.
  • When a lower-dimensional projection is required, ranking the null-space directions by the generalized eigenproblem (12) is reported to give consistently better retrieval performance than ranking by the other eigenproblems considered.
  • The whitening step used by UCSDA and OCSDA aligns the numerically computed null space of the intra-class scatter with the row space of the out-of-class scatter, and the regularization in ROCSDA restores a usable ranking of the projection vectors.
  • Combining null-space analysis with the cluster-based out-of-class scatter yields HNCSDA and HOCSDA, whose performance on face and scene datasets is comparable to or better than recent class-specific discriminant methods, with the number of clusters K as a tunable parameter that interpolates between binary LDA and CSDA.

Reading between the lines

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

  • The same null-space alignment logic could be exported to other discriminant-analysis settings that use regularized scatter matrices, potentially explaining why adding a small multiple of the identity often works well in practice.
  • Since the heterogeneous formulation splits the negative class into clusters, a natural testable extension is to use the within-cluster scatter as a regularizer or as an additional constraint, which could matter for open-set recognition where negatives are unlabeled but structurally organised.
  • The paper's observations about numerical conditioning suggest a general preprocessing rule for null-space methods: whiten the total scatter before extracting null spaces whenever the projection vectors must be ranked, and apply only a mild regularization to keep the eigenvalues informative.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. This paper proposes a null-space analysis for class-specific discriminant analysis (CSDA). It defines positive and negative scatter matrices Sp and Sn, derives the identity Nt = Np ∩ Nn for their null spaces, and then claims that when the total scatter St = Sp + Sn is full rank and the small-sample rank identity holds, the null space of Sp equals the row space of Sn. On this basis it formulates several algorithms: NCSDA (Algorithm 1), UCSDA/OCSDA/ROCSDA (Algorithm 2), and heterogeneous variants HNCSDA/HOCSDA (Algorithms 3–4). The methods are evaluated on seven datasets with average precision, and the heterogeneous variants are reported to be competitive with recent CSDA methods.

Significance. If the central subspace identity were true, the paper would provide a clean theoretical justification for extending null-space LDA ideas to class-specific problems, together with a principled ranking of projection directions. The paper also contains a detailed ablation of algorithmic steps, a useful discussion of numerical issues such as scatter-matrix symmetry and ill-conditioned St, and a broad experimental comparison on several standard datasets. However, the central identity is false, and the claimed optimality results and the justification of several eigenproblem choices rest on it. The theoretical contribution is therefore not sound as stated, although the empirical study may retain value as an algorithmic exploration.

major comments (4)
  1. [Section IV, after Eq. (5)] The implication 'When St is full rank, we have Nt = ∅, which means that Np = N⊥n' is false. A concrete counterexample satisfying the paper's rank assumptions is Sp = (1,0,-2)(1,0,-2)^T and Sn = diag(1,1,0) in R^3: rank(Sp) = 1, rank(Sn) = 2, St = Sp + Sn has full rank, and rank(St) = rank(Sp) + rank(Sn), exactly the N-1 ≤ r regime used before Eq. (5). Nevertheless null(Sp) = span(e2, (2,0,1)) while range(Sn) = span(e1, e2), so the vector (2,0,1) lies in null(Sp) but not in range(Sn). Consequently the statement that directions satisfying (3) also satisfy (2) is not valid. This invalidates the justification for Steps 2 and 3 of Algorithm 1, where the null space of St is removed in order to 'obtain Np = N⊥n'.
  2. [Section IV, Eq. (6)] The claim that the eigenvalues of St are the union of the nonzero eigenvalues of Sp and Sn is another consequence of the false equality and is false in the same example: the eigenvalues of St are 3 ± sqrt(5) and 1, while the nonzero eigenvalues of Sp and Sn are 5 and 1, 1. The subsequent discussion of numerical instability, including the observation that 'the null space of Sp and the row space of Sn are not properly aligned,' describes a real phenomenon, but the counterexample shows that the misalignment is structural rather than merely a numerical artifact of ill-conditioning.
  3. [Section IV-B, after Eq. (16)] The statement that NCSDA 'maximizes the criterion for d = rank(Sn) if Np = N⊥n' and that this can be achieved by removing the null space of St whenever rank(St) = rank(Sp) + rank(Sn) is unsupported, because the premise Np = N⊥n is false. The conclusion that NCSDA provides a solution to UCSDA/OCSDA whenever the constraints are satisfied therefore does not follow. The empirical success of the proposed variants may still be real, but it is not backed by the paper's theoretical analysis.
  4. [Section V-C1, Table III] The paper's own experiments illustrate the failure of the alignment assumption: solving (10) produces vectors that violate the null constraint A (for example, A = 335.74 on BU training data with full dimensionality), and the paper explains this by numerical instability. Given the counterexample above, large A values are expected even with exact arithmetic for some scatter-matrix pairs. The theory should be corrected rather than attributing the discrepancy to conditioning alone.
minor comments (4)
  1. [Section IV-C] The sentence 'rank(Sw) = rank(Snb) + rank(Snb)' appears to contain a typo; it should presumably be rank(Sn) = rank(Snw) + rank(Snb).
  2. [Table VI] The paper reports mean average precision averaged over five repetitions but does not report standard deviations or significance tests; adding variability measures would strengthen the comparisons, especially where differences between methods are small.
  3. [Section V-B] The regularization parameters μ = 10^-4, α = 10^-7, and the zero-eigenvalue threshold ε = 10^-6 are fixed; a brief sensitivity analysis would clarify how dependent the conclusions are on these choices.
  4. [Section IV-A, Algorithm 1] The argument that Step 4 has no effect uses the trace invariance tr(A^T B A) = tr(B) for orthogonal A, but when a subset of eigenvectors is selected for dimensionality reduction the matrix M is rectangular; the text should state this distinction explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the derivation is self-contained linear algebra; the disputed Np=N⊥n step is an invalid mathematical inference, not a reduction of the result to its inputs.

full rationale

The paper's derivation chain is not circular. Equation (5), Nt = Np ∩ Nn, follows from St = Sp + Sn together with positive semidefiniteness, not from the target result. The small-sample rank condition rank(St) = rank(Sp) + rank(Sn) is stated before it is used and follows from the standard rank identities for centered scatter matrices when N−1 ≤ r. The heterogeneous scatter decomposition Sn = Snw + Snb is re-derived in Section IV-C rather than imported as a black box, even though the formulation is credited to the authors' earlier work [21]. The comparison against [21] in Table VI is direct and the proposed methods sometimes underperform it, so the evaluation is not rigged. Hyperparameters d and K are set by cross-validation and μ, ε, α are fixed, not tuned to force conclusions. The paper's central theoretical step, 'When St is full rank, we have Nt = ∅, which means that Np = N⊥n,' is mathematically false: a 3x3 counterexample with Sp = c(1,0,−2)(1,0,−2)^T and Sn = diag(1,1,0) satisfies the paper's rank regime while null(Sp) ≠ range(Sn). However, this is a correctness error, not circularity: the false implication is not equivalent to an input by construction, and it is not a fitted parameter renamed as a prediction. The paper even observes in Section IV-A that the null space of Sp and the row space of Sn 'are not properly aligned' in practice, which is a validity limitation rather than a circular step. Minor self-citations [29] and [30] are background references and are not load-bearing for the proposed derivations. Overall, no circular step is present; the mathematical flaw should be evaluated as a correctness risk, not as circularity.

Assumptions & free parameters 6 free parameters · 7 assumptions · 0 invented entities

The algorithms rest on standard linear algebra plus a set of domain assumptions about the data (small sample size, linear independence after NPT, clusterable negative class). The experimental comparisons rest on several fixed or cross-validated hyperparameters rather than a parameter-free derivation.

free parameters (6)
  • subspace dimensionality d (1-25) = selected via 5-fold cross-validation
    For NCSDA, UCSDA, OCSDA, ROCSDA variants using reduced dimensionality; the reported performance gains depend on this choice.
  • number of negative-class clusters K = selected from {1,2,3,5,10} via 5-fold cross-validation
    Used in HNCSDA and HOCSDA to decompose the negative class scatter; determines the innate subspace dimensionality and the discriminant objective.
  • ROCSDA whitening regularization α = 1e-7
    Set to a value smaller than the eigenvalue threshold ε (1e-6) to demonstrate that regularized whitening improves projection ranking; no sensitivity analysis is provided.
  • scatter regularization μ = 1e-4
    Added to scatter matrices in generalized eigenproblems (11) and (12); fixed for all experiments.
  • zero-eigenvalue threshold ε = 1e-6
    Used to decide which eigenvalues are zero in solving the eigenproblems; affects the estimated null space.
  • RBF kernel width σ = sqrt(mean over training data of squared feature values)
    Kernel scaling set by a fixed rule rather than optimized; all experiments use this rule, so the results are conditional on it.
assumptions (7)
  • standard math Symmetric matrices have orthogonal eigenvectors and their nonzero eigenvalues are positive.
    Used throughout to justify eigenvector orthogonality and eigenvalue ranking (Section IV-A).
  • standard math For positive semi-definite matrices A and B, null(A+B) = null(A) ∩ null(B).
    Used to derive Eq. 5 in Section IV.
  • domain assumption After NPT preprocessing, the data dimensionality is r = rank(K) and the total scatter matrix St is full rank.
    The paper assumes this in Section IV to set r = N-1 and to make Nt empty.
  • domain assumption Training samples are linearly independent in the kernel feature space.
    Assumed before the rank equalities in Section IV; NPT discards zero eigenvalues to make this hold.
  • domain assumption Small sample size condition N - 1 <= r holds, so scatter matrices are singular and null spaces are nontrivial.
    The paper restricts the analysis to this regime; the proposed null-space methods are vacuous otherwise (Section IV).
  • domain assumption The negative class consists of K clusters and K-means can recover them.
    Used in HNCSDA and HOCSDA to compute between-cluster scatter Snb (Section IV-C).
  • ad hoc to paper Eigenproblem (12) is selected as Step 3 for NCSDA based on experimental performance.
    The paper does not derive this choice geometrically; it is chosen after experiments show it ranks projection vectors well (Section V-C1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Null Space Analysis for Class-Specific Discriminant Learning." pith.science (2026). https://pith.science/paper/7KBNJZBO

@misc{pith2026190804562,
  author       = {Pith},
  title        = {Pith review of: Null Space Analysis for Class-Specific Discriminant Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7KBNJZBO}},
  note         = {Machine review of arXiv:1908.04562}
}
read the original abstract

In this paper, we carry out null space analysis for Class-Specific Discriminant Analysis (CSDA) and formulate a number of solutions based on the analysis. We analyze both theoretically and experimentally the significance of each algorithmic step. The innate subspace dimensionality resulting from the proposed solutions is typically quite high and we discuss how the need for further dimensionality reduction changes the situation. Experimental evaluation of the proposed solutions shows that the straightforward extension of null space analysis approaches to the class-specific setting can outperform the standard CSDA method. Furthermore, by exploiting a recently proposed out-of-class scatter definition encoding the multi-modality of the negative class naturally appearing in class-specific problems, null space projections can lead to a performance comparable to or outperforming the most recent CSDA methods.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 43 canonical work pages

  1. [1]

    Face verification using c lient specific Fisher faces,

    Y . Kittler, J. Li, and J. Matas, “Face verification using c lient specific Fisher faces,” The Statistics of Directions, Shapes and Images , pp. 63– 66, 2000

    work page 2000

  2. [2]

    Activity based pers on identification using fuzzy representation and discriminant learning,

    A. Iosifidis, A. Tefas, and A. Pitas, “Activity based pers on identification using fuzzy representation and discriminant learning,” IEEE Transac- tions on Information F orensics and Security , vol. 7, no. 2, pp. 530–542, 2012

    work page 2012

  3. [3]

    R. Duda, P . Hart, and D. Stork, Pattern Classification, 2nd ed. Wiley- Interscience, 2000

    work page 2000

  4. [4]

    On the optimal class representation in Linear Discriminant Analysis,

    A. Iosifidis, A. Tefas, and I. Pitas, “On the optimal class representation in Linear Discriminant Analysis,” IEEE Transactions on Neural Networks and Learning Systems , vol. 24, no. 9, pp. 1491–1497, 2013. TABLE IV AVERAGE PRECISION FOR DIFFERENT UCSDA, OCSDA, AND ROCSDA VARIANTS (17) (18) (12) full D 1-25D full D 1-25D full D 1-25D JAFFE w/o wh. 0.465 0.4...

    work page 2013

  5. [5]

    Kernel reference di scriminant analysis,

    A. Iosifidis, A. Tefas, and I. Pitas, “Kernel reference di scriminant analysis,” Pattern Recognition Letters , vol. 49, pp. 85–91, 2014

    work page 2014

  6. [6]

    Class -specific ker- nel discriminant analysis for face verification,

    G. Goudelis, S. Zafeiriou, A. Tefas, and I. Pitas, “Class -specific ker- nel discriminant analysis for face verification,” IEEE Transactions on Information F orensics and Security, vol. 2, no. 3, pp. 570–587, 2007

    work page 2007

  7. [7]

    Regularized kernel discriminant analysis with a robust kernel for face r ecognition and verification,

    S. Zafeiriou, G. Tzimiropoulos, M. Petrou, and T. Statha ki, “Regularized kernel discriminant analysis with a robust kernel for face r ecognition and verification,” IEEE Transactions on Neural Networks , vol. 23, no. 3, pp. 526–534, 2012

    work page 2012

  8. [8]

    Class-specific kernel fusio n of multiple de- scriptors for face verification using multiscale binarized statistical image features,

    S. Arashloo and J. Kittler, “Class-specific kernel fusio n of multiple de- scriptors for face verification using multiscale binarized statistical image features,” IEEE Transactions on Information F orensics and Security , vol. 9, no. 12, pp. 2100–2109, 2014

    work page 2014

Show all 43 references
  1. [9]

    Class-specific refe rence discriminant analysis with application in human behavior analysis,

    A. Iosifidis, A. Tefas, and I. Pitas, “Class-specific refe rence discriminant analysis with application in human behavior analysis,” IEEE Transac- tions on Human-Machine Systems , vol. 45, no. 3, pp. 315–326, 2015

    2015

  2. [10]

    Class-specific n onlinear projec- tions using class-specific kernel spaces,

    A. Iosifidis, M. Gabbouj, and P . Pekki, “Class-specific n onlinear projec- tions using class-specific kernel spaces,” IEEE International Conference on Big Data Science and Engineering , pp. 1–8, 2015

    2015

  3. [11]

    Solving the small samp le size problem of lda,

    R. Huang, Q. Liu, H. Lu, and S. Ma, “Solving the small samp le size problem of lda,” in Object recognition supported by user interaction for service robots, vol. 3, Aug 2002, pp. 29–32

    2002

  4. [12]

    A new lda-based face recognition system which can solve the small sample size pro blem,

    F. Chen, H. Liao, M. Ko, J. Lin, and G. Y u, “A new lda-based face recognition system which can solve the small sample size pro blem,” 12 TABLE V EVALUATION OF CONSTRAINT A AND CRITERION B ON BU DATASET FOR OCSDA AND ROCSDA VARIANTS USING DIFFERENT APPROACHES FOR STEP 4 IN ALG...

    2000

  5. [13]

    Computational and theoretical anal ysis of null space and orthogonal linear discriminant analysis,

    J. Y e and T. Xiong, “Computational and theoretical anal ysis of null space and orthogonal linear discriminant analysis,” Journal of Machine Learning Research, vol. 7, pp. 1183–1204, 2006

    2006

  6. [14]

    Null space versus orthogonal linear discriminant analysis,

    J. Y e and T. Xiong, “Null space versus orthogonal linear discriminant analysis,” International Conference on Machine Learning , 2006

    2006

  7. [15]

    Kernel null space methods for novelty detection,

    P . Bodesheim, A. Freytag, E. Rodner, M. Kemmler, and J. D enzler, “Kernel null space methods for novelty detection,” IEEE Conference on Computer Vision and Pattern Recognition , 2013

    2013

  8. [16]

    A null space based one class k ernel fisher discriminant,

    F. Dufrenois and J. Noyer, “A null space based one class k ernel fisher discriminant,” International Joint Conference on Neural Networks , 2016

    2016

  9. [17]

    A theorem o n the uncorrelated optimal discriminant vectors,

    Z. Jin, J.-Y . Y ang, Z.-M. Tang, and Z.-S. Hu, “A theorem o n the uncorrelated optimal discriminant vectors,” Pattern Recognition, vol. 34, no. 10, pp. 2041 – 2047, 2001

    2001

  10. [18]

    Face recogniti on based on the uncorrelated discriminant transformation,

    Z. Jin, J.-Y . Y ang, Z.-S. Hu, and Z. Lou, “Face recogniti on based on the uncorrelated discriminant transformation,” Pattern Recognition, vol. 34, no. 7, pp. 1405 – 1416, 2001

    2001

  11. [19]

    Feature reductio n via gener- alized uncorrelated linear discriminant analysis,

    J. Y e, R. Janardan, Q. Li, and H. Park, “Feature reductio n via gener- alized uncorrelated linear discriminant analysis,” IEEE Transactions on Knowledge and Data Engineering , vol. 18, no. 10, pp. 1312–1322, 2006

    2006

  12. [20]

    Characterization of a family of algorithms for g eneralized discriminant analysis on undersampled problems,

    J. Y e, “Characterization of a family of algorithms for g eneralized discriminant analysis on undersampled problems,” Journal of Machine Learning Research, vol. 6, pp. 483–502, 2005

    2005

  13. [21]

    Probabilistic class-specific discrimin ant analysis,

    A. Iosifidis, “Probabilistic class-specific discrimin ant analysis,” arXiv:1812.05980, 2018

    2018 arXiv

  14. [22]

    Scholkpf and A

    B. Scholkpf and A. Smola, Learning with Kernels . MIT Press, 2001

    2001

  15. [23]

    Nonlinear Projection Trick in kernel methods : an alterna- tive to the kernel trick,

    N. Kwak, “Nonlinear Projection Trick in kernel methods : an alterna- tive to the kernel trick,” IEEE Transactions on Neural Networks and Learning Systems , vol. 24, no. 12, pp. 2113–2119, 2013

    2013

  16. [24]

    Nystr¨ om-based approxima te kernel sub- space learning,

    A. Iosifidis and M. Gabbouj, “Nystr¨ om-based approxima te kernel sub- space learning,” Pattern Recognition, vol. 57, pp. 190–197, 2016

    2016

  17. [25]

    Trace ratio problem revisi ted,

    Y . Jia, F. Nie, and C. Zhang, “Trace ratio problem revisi ted,” IEEE Transactions on Neural Networks , vol. 20, no. 4, pp. 729–735, 2009

    2009

  18. [26]

    Regularized discriminant analysis,

    J. Friedman, “Regularized discriminant analysis,” Journal of the Amer- ican Statistical Association , pp. 165–175, 1989

    1989

  19. [27]

    An optimizatio n criterion for generalized discriminant analysis on undersampled pro blems,

    J. Y e, R. Janardan, C. Park, and H. Park, “An optimizatio n criterion for generalized discriminant analysis on undersampled pro blems,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 26, no. 8, pp. 982–994, Aug 2004

    2004

  20. [28]

    Spectral regression for effici ent regularized subspace learning,

    D. Cai, X. He, and J. Han, “Spectral regression for effici ent regularized subspace learning,” International Conference on Computer Vision , 2007

    2007

  21. [29]

    Scaling up class-specific k ernel discrim- inant analysis for large-scale face verification,

    A. Iosifidis and M. Gabbouj, “Scaling up class-specific k ernel discrim- inant analysis for large-scale face verification,” IEEE Transactions on Information F orensics and Security , vol. 11, no. 11, pp. 2453–2465, 2016

    2016

  22. [30]

    Class-specific kernel disc riminant anal- ysis revisited: Further analysis and extensions,

    A. Iosifidis and M. Gabbouj, “Class-specific kernel disc riminant anal- ysis revisited: Further analysis and extensions,” IEEE Transactions on Cybernetics, vol. 47, no. 12, pp. 4485–4496, 2017

    2017

  23. [31]

    Eigenanalysis of symmetr izable matrix products: a result with statistical applications,

    N. Wermuth and H. Riissmann, “Eigenanalysis of symmetr izable matrix products: a result with statistical applications,” Scandinavian Journal of Statistics, vol. 20, pp. 361–367, 1993

    1993

  24. [32]

    Null foley-samm on transform,

    Y . Guo, L. Wu, H. Lu, Z. Feng, and X. Xue, “Null foley-samm on transform,” Pattern Recognition, vol. 39, no. 11, pp. 2248–2251, 2006

    2006

  25. [33]

    Parlett, The Symmetric Eigenvalue Problem

    B. Parlett, The Symmetric Eigenvalue Problem . Society for Industrial and Applied Mathematics, 1998

    1998

  26. [34]

    Coding facial expressions with gabor wavelets,

    M. Lyons, S. Akamatsu, M. Kamachi, and J. Gyoba, “Coding facial expressions with gabor wavelets,” in IEEE International Conference on Automatic Face and Gesture Recognition , April 1998, pp. 200–205

    1998

  27. [35]

    Comprehensive d atabase for facial expression analysis,

    T. Kanade, J. F. Cohn, and Yingli Tian, “Comprehensive d atabase for facial expression analysis,” in IEEE International Conference on Automatic Face and Gesture Recognition , March 2000, pp. 46–53

    2000

  28. [36]

    A 3d facial expression database for facial behavior research,

    L. Yin, X. Wei, Y . Sun, J. Wang, and M. Rosato, “A 3d facial expression database for facial behavior research,” in IEEE International Conference on Automatic Face and Gesture Recognition , April 2006, pp. 211–216

    2006

  29. [37]

    Parameterisation of a st ochastic model for human face identification,

    F. S. Samaria and A. C. Harter, “Parameterisation of a st ochastic model for human face identification,” in IEEE W orkshop on Applications of Computer Vision, Dec 1994, pp. 138–142

    1994

  30. [38]

    The AR-face database,

    A. Martinez and R. Benavente, “The AR-face database,” T ech. Rep., 1998

    1998

  31. [39]

    Acquiring linear sub spaces for face recognition under variable lighting,

    K.-C. Lee, J. Ho, and D. Kriegman, “Acquiring linear sub spaces for face recognition under variable lighting,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 27, no. 5, pp. 684–698, 2005

    2005

  32. [40]

    Beyond bags of fea tures: Spatial pyramid matching for recognizing natural scene categories ,

    S. Lazebnik, C. Schmid, and J. Ponce, “Beyond bags of fea tures: Spatial pyramid matching for recognizing natural scene categories ,” in IEEE Conference on Computer Vision and Pattern Recognition , vol. 2, June 2006, pp. 2169–2178

    2006

  33. [41]

    The Pascal Visual Object Classes (VOC) Challenge,

    M. Everingham, L. V an Gool, C. Williams, J. Winn, and A. Z isserman, “The Pascal Visual Object Classes (VOC) Challenge,” International Journal of Computer Vision , vol. 88, no. 2, pp. 303–338, Jun 2010

    2010

  34. [42]

    An algorithm for generali zed matrix eigenvalue problems,

    C. B. Moler and G. W. Stewart, “An algorithm for generali zed matrix eigenvalue problems,” SIAM Journal on Numerical Analysis , vol. 10, no. 2, pp. 241–256, 1973

    1973

  35. [43]

    Analysis of the ch olesky method with iterative refinement for solving the symmetric definite generalized eigenproblem,

    P . Davies, N. Higham, and F. Tisseur, “Analysis of the ch olesky method with iterative refinement for solving the symmetric definite generalized eigenproblem,” SIAM Journal on Matrix Analysis and Applications , vol. 23, no. 2, pp. 472–493, 2001

    2001

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.