Pith. sign in

REVIEW 5 major objections 6 minor 39 references

Noise-Resistant Label Reconstruction Feature Selection for Partial Multi-Label Learning

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · 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 three-stage method for partial multi-label learning, PML-FSMIR, cleans the noisy candidate-label matrix using label-to-label mutual information, learns feature weights without reducing the data dimension, and then…

desk verdict PML-FSMIR is a genuinely new three-stage pipeline with large reported F1 gains, but the paper's update equations don't match its stated objective and its headline ranking claim is misstated — both fixable, but not ignorable. read the letter →

arxiv 2506.04669 v1 pith:T2ON66VH submitted 2025-06-05 cs.LG

Wanfu Gao , Hanlin Pan , Qingqi Han , Kunpeng Liu This is my paper
classification cs.LG
keywords partialmulti-labellearningfeatureselectionmutualinformationlabelnoisereconstructionconnectivitypositiveidentificationsparse
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 proposes a feature-selection method for partial multi-label learning, the setting where each example carries a candidate label set and at least one label is correct while the rest may be noise. The method, PML-FSMIR, uses mutual information between labels to rebuild the noisy label matrix, learns feature weights without collapsing the data to a low-rank space, and finally reweights features toward labels that are strongly connected to many other labels. The paper reports that on eight benchmark datasets and five evaluation metrics this procedure ranks first in every combination except Coverage on CHD49 and Average Precision on LLOGF, where it ranks second. A sympathetic reader would take the paper's claim to be that label-relationship structure is stable enough under label noise to serve as the foundation for both label disambiguation and reliable feature selection.

What carries the argument

The central objects are the two mutual-information matrices: $Z_{ij} = I(y_{:i}, y_{:j})$ computed from the observed candidate labels, and $Z'_{ij} = I(T_{:i}, T_{:j})$ computed from the reconstructed label matrix. The first matrix enters $T = (Y Z) \circ \mathrm{sign}(Y)$, reweighting each candidate label by the summed mutual information it shares with the other candidate labels in that sample; this is the noise-resistance mechanism, since an isolated candidate label with weak associations is downweighted as likely noise. The second matrix supplies the graph-Laplacian regularizer during weight learning and performs the final update $W \leftarrow W Z'$, which shifts weight toward features that identify labels with many strong connections. The reformed low-rank assumption keeps $X$ at its original dimension by using $UV$ as a denoised reconstruction of $X$ rather than as a replacement for it.

What would settle it

Construct synthetic partial multi-label datasets with known ground-truth labels, then add false candidate labels at several noise levels (say 10%, 30%, 50%, and 70%) and run PML-FSMIR on each; if its ranking loss or F1 degrades faster than a baseline that ignores label structure, or if the reconstructed label matrix $T$ disagrees with the known clean labels, the noise-resistance premise fails. The paper only tests one fixed 20% noise setting and never compares the reconstructed labels against a clean-label ground truth, so this check would settle the claim.

Watch

Extended reading notes

Core claim

PML-FSMIR claims that two collective properties of labels — noise-resistant label relationships and label connectivity — are sufficient for feature selection in partial multi-label learning without relying on the low-rank assumption used by earlier PML methods. The paper constructs a label-reconstruction matrix $T = (Y Z) \circ \mathrm{sign}(Y)$ from the candidate-label matrix $Y$ and the pairwise label mutual-information matrix $Z$, so a candidate label is treated as reliable when the other candidate labels in the same sample are strongly associated with it. The learning objective then combines $\|UVW - T\|_F^2$, $\alpha\|X - UV\|_F^2$, a manifold regularizer built from the mutual-information graph of $T$, and a $\gamma\|W\|_{2,1}$ term for feature selection, with $U$ and $V$ reconstructing $X$ while keeping its full dimension. In the final stage the learned weight matrix is replaced by $W Z'$, which magnifies weights for features that discriminate representative, well-connected labels. The paper argues that this ordering of operations removes label noise before learning, avoids information loss from low-rank projection, and concentrates the selected features on labels that are important by connectivity.

Load-bearing premise

The load-bearing assumption is that the label-to-label association statistics computed from the observed candidate labels are still trustworthy, so that a candidate label surrounded by associated candidate labels is likely correct and an isolated one is likely noise; if the label noise is strong enough to distort those statistics, the reconstruction would amplify the noise instead of removing it.

Editorial extensions

If this is right

  • Ranked feature subsets produced by the method should improve SVM classification on partial multi-label data across audio, music, text, image, chemistry, biology, and medicine benchmarks.
  • The third-stage weight reconstruction is what makes positive-label identification possible: without it the paper reports that the method identifies no positive labels on six of eight datasets, and with it the method identifies positive labels on all eight.
  • Because the method keeps the original feature dimension rather than replacing $X$ by a low-rank factor, it should retain high-order structural information that low-rank PML methods discard.
  • The method's ranking behavior is stable to the three regularization parameters $\alpha$, $\beta$, and $\gamma$ across the tested range from 0.001 to 1000.

Reading between the lines

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

  • If the co-occurrence-based reconstruction is as effective as reported, the natural next test is to vary the noise level instead of keeping it fixed at 20%, since the reconstruction's premise is that label co-occurrence statistics survive moderate noise.
  • The same mutual-information reweighting could transplant directly into deep partial multi-label classifiers, where the label graph is often available but label noise is handled by separate modules.
  • The weight-reconstruction rule implies a testable ranking hypothesis: features with high weight on high-degree labels should transfer to better detection of rare positive labels, which could be checked on a dataset with explicit label-degree annotations.
  • The reported F1 improvements are large in absolute terms, so a paired significance test across the ten-fold cross-validation splits would clarify whether the gap is stable rather than driven by a few folds.
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

5 major / 6 minor

Summary. The paper proposes PML-FSMIR, a three-stage feature selection method for partial multi-label learning (PML). Stage 1 reconstructs the candidate label matrix using a mutual-information matrix to suppress label noise (Eq. 2). Stage 2 learns a feature weight matrix through a 'reformed low-rank' objective that includes a graph-Laplacian regularization term on the weight matrix (Eq. 7). Stage 3 reconstructs the weight matrix using label-connectivity information to emphasize features that are important for representative labels (Eq. 12). The method is evaluated on eight benchmark datasets against eight existing methods, and the paper reports ablation studies and parameter sensitivity analyses.

Significance. The general idea of using stable label relationships, via mutual information, for noise-resistant feature selection in PML is interesting and the authors provide a code repository and extensive experiments. If the method worked exactly as claimed, it would be a useful contribution to PML feature selection. However, the manuscript in its current form contains load-bearing inconsistencies: the optimization update for W does not match the stated objective, the claim of breaking free of the low-rank assumption is contradicted by the model itself, and the central empirical superiority claim is contradicted by the paper's own tables. These issues prevent a reliable assessment of the method's contribution.

major comments (5)
  1. [3.2, Eq. (11)] The multiplicative update for W omits the graph Laplacian term that is central to the objective in Eq. (7). Differentiating β Tr(W L_T W^T) with respect to W yields a term proportional to L_T W (or W L_T), but the denominator in Eq. (11) contains β W instead. Furthermore, the parameter δ in Eq. (11) is never defined; presumably it should be related to γ from the l2,1-norm relaxation, but this is not stated. Consequently, the iterative procedure described in Algorithm 1 does not minimize the stated objective, and the reported experimental results cannot be attributed to the method as specified.
  2. [3.2, Eq. (4); Introduction and Contributions] The paper repeatedly claims that the method 'breaks free of the low-rank assumption' and 'preserves the dimensions of the sample space.' This is contradicted by Eq. (4), which contains α||X - U V||^2_F with U ∈ R^{n×k} and V ∈ R^{k×d}. This is exactly a rank-k approximation of X, and the first term uses U V, which is a low-rank product. The formulation therefore still relies on a low-rank factorization of the feature matrix; the claim of avoiding the low-rank assumption is a conceptual overstatement and is not supported by the model.
  3. [4.3, Tables 3 and 4] The key empirical claim that PML-FSMIR 'ranks first in all cases except Coverage on CHD49 and Average Precision on LLOGF, while in these two cases PML-FSMIR all ranks second' is false. In Table 3 (Coverage, lower is better), on LLOGF PML-FSMIR has 0.53 while PML-LC has 0.51 and PML-FP has 0.48, so PML-FSMIR ranks third, not second. On CHD49 in the same table, PML-FSMIR has 0.50 while PML-FSSO and DRMFS both have 0.49, again ranking third. The prose also omits Coverage on LLOGF as an additional case where the method is not first. Because this ranking statement is the paper's central evidence of superiority, the quantitative support for the main contribution is lacking.
  4. [3.1, Eq. (3)] The expression T_{ij} = y_{ij} Σ_{k=1}^q y_{ik} I(y_{ij}, y_{ik}) uses the notation I(y_{ij}, y_{ik}), which is undefined: mutual information is a function of two random variables (label columns), not of individual entries. The intended meaning is presumably the mutual-information matrix entry Z_{jk} = I(y_{:j}, y_{:k}). This notational error makes the derivation of the label reconstruction matrix difficult to follow and should be corrected.
  5. [3.1 and 4.5] The noise-resistance premise of the first stage is not adequately tested. The mutual information matrix Z in Eq. (1) is computed from the observed candidate label matrix Y, which already contains noise, and the reconstruction T in Eq. (2) assumes that label co-occurrence statistics are stable under noise. The experiments use only a fixed 20% noise level and provide no validation against clean ground-truth labels. Without varying the noise level or otherwise demonstrating that the estimated Z preserves the true label structure under corruption, the claimed 'noise-resistant' property of the label reconstruction stage is not established.
minor comments (6)
  1. [4.3 and Tables 5-6] The terms 'Marco-F1' and 'Mirco-F1' appear throughout; these should be 'Macro-F1' and 'Micro-F1'.
  2. [4.4, Figure 5] The text states that Figure 5 shows parameter sensitivity on LLOGF in terms of Ranking Loss, but the figure caption says 'Parameter sensitivity studies on the CAL in terms of Coverage.' The x-axis labels in the figure are also garbled ('0.001 10008% 20%').
  3. [4.3] The text refers to 'Figure ??' when discussing the per-dataset curves; this should be a concrete figure number, presumably Figure 4.
  4. [Algorithm 1, line 4] The loop condition 'while not coverage do' should be 'while not converged do'.
  5. [3.2, Eqs. (6)-(8)] The graph Laplacian regularization term is written inconsistently across equations: 'β Tr(W )LT (W )T' in Eq. (6), 'β Tr(W )LT (W )T' in Eq. (7), and 'β TrW L_T W^T' in Eq. (8). The intended form is Tr(W L_T W^T), and the notation should be uniform.
  6. [3.2, after Eq. (8)] The citation of Beck and Teboulle [2009] for the multiplicative gradient descent strategy is inaccurate; that reference introduces FISTA, not multiplicative update rules. A more appropriate reference would be to multiplicative non-negative matrix factorization updates.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: PML-FSMIR is an empirical optimization-based feature-selection method whose steps do not reduce by construction to their inputs; the paper's reported ranking claim is internally inconsistent with Table 3, but that is a correctness issue, not circularity.

full rationale

No load-bearing circular step was found. The three stages are explicit algorithmic operations: the label reconstruction matrix T = (Y Z) sign(Y) in Eq. (2) is a preprocessing transform of the observed noisy label matrix, not a fitted parameter disguised as a prediction. The mutual-information matrices Z and Z' are computed from observed data, and the objective in Eq. (7) optimizes U, V, and W against the reconstructed T with regularization parameters alpha, beta, and gamma; the final feature ranking is the row-wise l2-norm of the reconstructed weight matrix after Eq. (12). This is the method's own selection criterion, not a claim that ground-truth labels were derived from the inputs by construction. The ablation experiments compare versions of the method on the same benchmark data and are internal consistency checks rather than circular derivations. The self-citations (for example, Gao et al. 2018 and Hu et al. 2020) appear only as related work and are not load-bearing uniqueness theorems or unverified premises. The main caveat is that the paper's Section 4.3 claim that PML-FSMIR 'ranks first in all cases except Coverage on CHD 49 and Average Precision on LLOG F, while in these two cases PML-FSMIR all ranks second' is contradicted by Table 3: for Coverage on LLOGF, PML-FSMIR (0.53) is worse than PML-LC (0.51) and PML-FP (0.48), and on CHD49 (0.50) it is no better than PML-FSSO and DRMFS (0.49). This undermines the reported superiority claim, and the label-reconstruction premise is validated only on the same noisy benchmarks at a fixed 20% noise level. These are validity and verification concerns, not circularity; therefore the circularity score is 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim rests on four unproved modeling assumptions and four tunable hyperparameters, including the unresolved rank k. No new entities are introduced.

free parameters (4)
  • alpha = tuned over {0.001, 0.01, 0.1, 1, 10, 100, 1000}; no final value reported
    Regularization coefficient for the feature reconstruction term α∥X - U V∥²_F in Eq. 4. Sensitivity is shown only on one dataset.
  • beta = tuned over {0.001, 0.01, 0.1, 1, 10, 100, 1000}; no final value reported
    Regularization coefficient for the graph Laplacian term β Tr(W L_T W^T) in Eq. 6 and Eq. 7.
  • gamma = tuned over {0.001, 0.01, 0.1, 1, 10, 100, 1000}; no final value reported
    Regularization coefficient for the l2,1 sparsity term γ∥W∥_{2,1} in Eq. 7.
  • k = not specified
    Rank of the factorization U ∈ R^{n×k}, V ∈ R^{k×d} in Eq. 4. It is not listed in Algorithm 1's input and no selection procedure is given, so this is an undisclosed free parameter.
assumptions (4)
  • domain assumption The mutual information matrix Z computed from the observed noisy candidate label matrix Y reflects true label relationship structure (noise-resistance); T=(Y Z)∘sign(Y) therefore increases values of true positive labels and suppresses false positives.
    Section 3.1, Eq. 2. This is the core premise of stage 1 and is not verified against varying noise levels or a clean-label reference.
  • domain assumption Labels with more and stronger connections in the label mutual information graph are representative; features scoring highly for them improve positive-label identification.
    Section 3.3, Eq. 12. No external validation that graph degree corresponds to label importance; ablation only compares with and without the stage.
  • ad hoc to paper A rank-k factorization U V of X removes noise and redundancy while preserving high-dimensional structure, despite projecting X onto a rank-k subspace.
    Section 3.2, Eq. 4. 'Reformed low-rank assumption' still imposes rank ≤ k; the paper's claim to break free of low-rank assumptions is contradicted by this term.
  • standard math Multiplicative updates (Eqs. 9 to 11) converge to a stationary point of Eq. 7.
    Invoked but not proven; update for W in Eq. 11 does not include the graph Laplacian term, so no convergence to the stated objective is established.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Noise-Resistant Label Reconstruction Feature Selection for Partial Multi-Label Learning." pith.science (2026). https://pith.science/paper/T2ON66VH

@misc{pith2026250604669,
  author       = {Pith},
  title        = {Pith review of: Noise-Resistant Label Reconstruction Feature Selection for Partial Multi-Label Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T2ON66VH}},
  note         = {Machine review of arXiv:2506.04669}
}
read the original abstract

The "Curse of dimensionality" is prevalent across various data patterns, which increases the risk of model overfitting and leads to a decline in model classification performance. However, few studies have focused on this issue in Partial Multi-label Learning (PML), where each sample is associated with a set of candidate labels, at least one of which is correct. Existing PML methods addressing this problem are mainly based on the low-rank assumption. However, low-rank assumption is difficult to be satisfied in practical situations and may lead to loss of high-dimensional information. Furthermore, we find that existing methods have poor ability to identify positive labels, which is important in real-world scenarios. In this paper, a PML feature selection method is proposed considering two important characteristics of dataset: label relationship's noise-resistance and label connectivity. Our proposed method utilizes label relationship's noise-resistance to disambiguate labels. Then the learning process is designed through the reformed low-rank assumption. Finally, representative labels are found through label connectivity, and the weight matrix is reconstructed to select features with strong identification ability to these labels. The experimental results on benchmark datasets demonstrate the superiority of the proposed method.

Figures

Figures reproduced from arXiv: 2506.04669 by the authors.

Figure 1
Figure 1. An example of PML. Among the candidate set of seven [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Illustration of PML-FSMIR. In the first stage, label matrix is reconstructed with mutual information matrix to get the reliability of [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The weight matrix W is reconstructed through the mu￾tual information matrix Z ′ representing the label connection, and the correct feature weight is obtained. Q is a diagonal matrix: Qii = 1 2 √ WT i Wi+ϵ ,(ϵ → 0). The objective function can be rewritten as: Θ(U, V, W) = Tr (UV W − T) T (UV W − T)  + α Tr (X − UV ) T (X − UV )  + β Tr W LT WT + γ Tr WT QW . (8) Multiplicative gradient descent strategy is adopted … view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Nine methods on Birds in terms of Marco-F1, Average Precision, Ranking Loss and Coverage.. 4.3 Results The tables below show the detail of the experiment result, for all datasets except Water, we select specified number of fea￾tures according to the importance as desce…
Figure 7
Figure 7. Figure 7: The results of the comparison experiment between the first [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 5
Figure 5. Figure 5: Parameter sensitivity studies on the CAL in terms of Cov￾erage. Bi CA CH Co LL Sl Wa Ye 0.0 0.2 0.4 Stage 1,2 Stage 2 (a) Ranking Loss Bi CA CH Co LL Sl Wa Ye 0.0 0.2 0.4 0.6 0.8 Stage 1,2 Stage 2 (b) Average Precision [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The results of the comparison experiment between the first [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages

  1. [1]

    A fast iterative shrinkage-thresholding algorithm for lin- ear inverse problems

    [Beck and Teboulle, 2009] Amir Beck and Marc Teboulle. A fast iterative shrinkage-thresholding algorithm for lin- ear inverse problems. SIAM journal on imaging sciences, 2(1):183–202,

    work page 2009

  2. [13]

    Partial multi-label learning via multi-subspace represen- tation

    [Li et al., 2021] Ziwei Li, Gengyu Lyu, and Songhe Feng. Partial multi-label learning via multi-subspace represen- tation. In Proceedings of the Twenty-Ninth International Conference on International Joint Conferences on Artifi- cial Intelligence, pages 2612–2618,

    work page 2021

  3. [14]

    Multi-label feature selection with high-sparse personalized and low-redundancy shared common features.Information Processing & Management, 61(3):103633,

    [Li et al., 2024] Yonghao Li, Liang Hu, and Wanfu Gao. Multi-label feature selection with high-sparse personalized and low-redundancy shared common features.Information Processing & Management, 61(3):103633,

    work page 2024

  4. [18]

    The emerging trends of multi-label learning

    [Liu et al., 2021] Weiwei Liu, Haobo Wang, Xiaobo Shen, and Ivor W Tsang. The emerging trends of multi-label learning. IEEE transactions on pattern analysis and ma- chine intelligence, 44(11):7955–7974,

    work page 2021

  5. [19]

    Efficient and robust feature selection via joint l2, 1-norms minimization

    [Nie et al., 2010] Feiping Nie, Heng Huang, Xiao Cai, and Chris Ding. Efficient and robust feature selection via joint l2, 1-norms minimization. Advances in neural information processing systems, 23,

    work page 2010

  6. [22]

    Random forest

    [Rigatti, 2017] Steven J Rigatti. Random forest. Journal of Insurance Medicine, 47(1):31–39,

    work page 2017

  7. [25]

    Partial multi-label learning by low-rank and sparse decomposition

    [Sun et al., 2019] Lijuan Sun, Songhe Feng, Tao Wang, Con- gyan Lang, and Yi Jin. Partial multi-label learning by low-rank and sparse decomposition. In Proceedings of the AAAI conference on artificial intelligence, volume 33, pages 5016–5023,

    work page 2019

  8. [26]

    Semantic annota- tion and retrieval of music and sound effects

    [Turnbull et al., 2008] Douglas Turnbull, Luke Barrington, David Torres, and Gert Lanckriet. Semantic annota- tion and retrieval of music and sound effects. IEEE Transactions on Audio, Speech, and Language Processing, 16(2):467–476,

    work page 2008

Show all 39 references
  1. [28]

    Deep multi-view subspace clustering with unified and dis- criminative learning

    [Wang et al., 2021] Qianqian Wang, Jiafeng Cheng, Quanxue Gao, Guoshuai Zhao, and Licheng Jiao. Deep multi-view subspace clustering with unified and dis- criminative learning. IEEE Transactions on Multimedia , 23:3483–3493,

    2021

  2. [29]

    Partial multi-label learning

    [Xie and Huang, 2018] Ming-Kun Xie and Sheng-Jun Huang. Partial multi-label learning. In Proceedings of the AAAI conference on artificial intelligence, volume 32,

    2018

  3. [30]

    Partial multi-label learning with noisy label identification

    [Xie and Huang, 2021] Ming-Kun Xie and Sheng-Jun Huang. Partial multi-label learning with noisy label identification. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(7):3676–3687,

    2021

  4. [31]

    Robust extreme multi-label learning

    [Xu et al., 2016] Chang Xu, Dacheng Tao, and Chao Xu. Robust extreme multi-label learning. InProceedings of the 22nd ACM SIGKDD international conference on knowl- edge discovery and data mining, pages 1275–1284,

    2016

  5. [32]

    Partial multi-label learning with label distribution

    [Xu et al., 2020] Ning Xu, Yun-Peng Liu, and Xin Geng. Partial multi-label learning with label distribution. In Pro- ceedings of the AAAI conference on artificial intelligence, volume 34, pages 6510–6517,

    2020

  6. [33]

    Multi-label sentiment analysis on 100 languages with dynamic weighting for label imbalance

    [Yilmaz et al., 2021] Selim F Yilmaz, E Batuhan Kaynak, Aykut Koc ¸, Hamdi Dibeklio ˘glu, and Suleyman Serdar Kozat. Multi-label sentiment analysis on 100 languages with dynamic weighting for label imbalance. IEEE Trans- actions on Neural Networks and Learning Systems,

    2021

  7. [34]

    Feature-induced partial multi-label learning

    [Yu et al., 2018] Guoxian Yu, Xia Chen, Carlotta Domeni- coni, Jun Wang, Zhao Li, Zili Zhang, and Xindong Wu. Feature-induced partial multi-label learning. In2018 IEEE international conference on data mining (ICDM) , pages 1398–1403. IEEE,

    2018

  8. [35]

    Partial multi-label learning with label and feature collaboration

    [Yu et al., 2020] Tingting Yu, Guoxian Yu, Jun Wang, and Maozu Guo. Partial multi-label learning with label and feature collaboration. In Database Systems for Advanced Applications: 25th International Conference, DASFAA 2020, Jeju, South Korea, September 24–27, 2020, Pro- ceed...

    2020

  9. [36]

    Partial multi-label learning via credible label elicita- tion

    [Zhang and Fang, 2020] Min-Ling Zhang and Jun-Peng Fang. Partial multi-label learning via credible label elicita- tion. IEEE Transactions on Pattern Analysis and Machine Intelligence, 43(10):3587–3599,

    2020

  10. [37]

    Feature relevance term variation for multi-label feature selection

    [Zhang and Gao, 2021] Ping Zhang and Wanfu Gao. Feature relevance term variation for multi-label feature selection. Applied Intelligence, 51:5095–5110,

    2021

  11. [38]

    Ml-knn: A lazy learning approach to multi-label learning

    [Zhang and Zhou, 2007] Min-Ling Zhang and Zhi-Hua Zhou. Ml-knn: A lazy learning approach to multi-label learning. Pattern recognition, 40(7):2038–2048,

    2007

  12. [39]

    Feature selection based on mutual infor- mation with correlation coefficient

    [Zhou et al., 2022] Hongfang Zhou, Xiqian Wang, and Rourou Zhu. Feature selection based on mutual infor- mation with correlation coefficient. Applied intelligence, 52(5):5457–5474, 2022

    2022

  13. [1948]

    Symptom selection for multi-label data of inquiry diagnosis in traditional chinese medicine

    [Shao et al., 2013] Huan Shao, GuoZheng Li, GuoPing Liu, and YiQin Wang. Symptom selection for multi-label data of inquiry diagnosis in traditional chinese medicine. Sci- ence China Information Sciences, 56:1–13,

    2013

  14. [1999]

    The 9th annual mlsp compe- tition: New methods for acoustic classification of multiple simultaneous bird species in a noisy environment

    [Briggs et al., 2013] Forrest Briggs, Yonghong Huang, Ra- viv Raich, Konstantinos Eftaxias, Zhong Lei, William Cukierski, Sarah Frey Hadley, Adam Hadley, Matthew Betts, Xiaoli Z Fern, et al. The 9th annual mlsp compe- tition: New methods for acoustic classification of multiple...

    2013

  15. [2001]

    Class-specific mutual information variation for feature se- lection

    [Gao et al., 2018] Wanfu Gao, Liang Hu, and Ping Zhang. Class-specific mutual information variation for feature se- lection. Pattern Recognition, 79:328–339,

    2018

  16. [2002]

    A kernel method for multi-labelled classification

    [Elisseeff and Weston, 2001] Andr´e Elisseeff and Jason We- ston. A kernel method for multi-labelled classification. Advances in neural information processing systems , 14,

    2001

  17. [2006]

    Low-rank multi-view learning in matrix completion for multi-label image classification

    [Liu et al., 2015] Meng Liu, Yong Luo, Dacheng Tao, Chao Xu, and Yonggang Wen. Low-rank multi-view learning in matrix completion for multi-label image classification. In Proceedings of the AAAI conference on artificial intel- ligence, volume 29,

    2015

  18. [2008]

    Discrimina- tive and correlative partial multi-label learning

    [Wang et al., 2019] Haobo Wang, Weiwei Liu, Yang Zhao, Chen Zhang, Tianlei Hu, and Gang Chen. Discrimina- tive and correlative partial multi-label learning. In IJCAI, pages 3691–3697,

    2019

  19. [2009]

    Simultaneous prediction of multiple chemical parameters of river water quality with tilde

    [Blockeel et al., 1999] Hendrik Blockeel, Sa ˇso D ˇzeroski, and Jasna Grbovi ´c. Simultaneous prediction of multiple chemical parameters of river water quality with tilde. In European Conference on Principles of Data Mining and Knowledge Discovery, pages 32–40. Springer,

    1999

  20. [2010]

    Graph-based multi-label disease prediction model learning from medi- cal data and domain knowledge

    [Pham et al., 2022] Thuan Pham, Xiaohui Tao, Ji Zhang, Jianming Yong, Yuefeng Li, and Haoran Xie. Graph-based multi-label disease prediction model learning from medi- cal data and domain knowledge. Knowledge-based sys- tems, 235:107662,

    2022

  21. [2013]

    Survey on svm and their application in image classification

    [Chandra and Bedi, 2021] Mayank Arya Chandra and SS Bedi. Survey on svm and their application in image classification. International Journal of Information Technology, 13(5):1–11,

    2021

  22. [2015]

    Semi- supervised multi-label learning by constrained non- negative matrix factorization

    [Liu et al., 2006] Yi Liu, Rong Jin, and Liu Yang. Semi- supervised multi-label learning by constrained non- negative matrix factorization. In AAAi, volume 6, pages 421–426,

    2006

  23. [2016]

    Recov- ering accurate labeling information from partially valid data for effective multi-label learning

    [Li and Wang, 2020] Ximing Li and Yang Wang. Recov- ering accurate labeling information from partially valid data for effective multi-label learning. arXiv preprint arXiv:2006.11488,

    2020 arXiv

  24. [2017]

    A mathematical theory of communication

    [Shannon, 1948] Claude Elwood Shannon. A mathematical theory of communication. The Bell system technical jour- nal, 27(3):379–423,

    1948

  25. [2018]

    A unified low-order information-theoretic feature selection framework for multi-label learning

    [Gao et al., 2023] Wanfu Gao, Pingting Hao, Yang Wu, and Ping Zhang. A unified low-order information-theoretic feature selection framework for multi-label learning. Pat- tern Recognition, 134:109111,

    2023

  26. [2019]

    Object recogni- tion as machine translation: Learning a lexicon for a fixed image vocabulary

    [Duygulu et al., 2002] Pinar Duygulu, Kobus Barnard, Joao FG de Freitas, and David A Forsyth. Object recogni- tion as machine translation: Learning a lexicon for a fixed image vocabulary. In Computer Vision—ECCV 2002: 7th European Conference on Computer Vision Copenhagen, Denm...

    2002

  27. [2020]

    Partial multi-label feature selection via subspace optimiza- tion

    [Hao et al., 2023] Pingting Hao, Liang Hu, and Wanfu Gao. Partial multi-label feature selection via subspace optimiza- tion. Information Sciences, 648:119556,

    2023

  28. [2021]

    Learning a deep convnet for multi-label classification with partial labels

    [Durand et al., 2019] Thibaut Durand, Nazanin Mehrasa, and Greg Mori. Learning a deep convnet for multi-label classification with partial labels. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 647–657,

    2019

  29. [2022]

    Scalable multi-label classification

    [Read, 2010] Jesse Read. Scalable multi-label classification. PhD thesis, University of Waikato,

    2010

  30. [2023]

    Distributed multi-label feature selection using individual mutual information mea- sures

    [Gonzalez-Lopez et al., 2020] Jorge Gonzalez-Lopez, Se- basti´an Ventura, and Alberto Cano. Distributed multi-label feature selection using individual mutual information mea- sures. Knowledge-Based Systems, 188:105052,

    2020

  31. [2024]

    Multi-label feature selection based on max-dependency and min-redundancy

    [Lin et al., 2015] Yaojin Lin, Qinghua Hu, Jinghua Liu, and Jie Duan. Multi-label feature selection based on max-dependency and min-redundancy. Neurocomputing, 168:92–103,

    2015

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.