Pith. sign in

REVIEW 3 major objections 6 minor 47 references

How Effective Can Dropout Be in Multiple Instance Learning ?

T0 review · 3 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read This paper argues that removing the top-ranked instances from a bag during MIL training—the opposite of what intuition suggests—consistently lowers gradient error, flattens the loss landscape, and improves classification, and it packages…

desk verdict The top-k instance-drop observation is genuinely worth testing, but the main tables conflate MIL-Dropout with an added feature extractor, so the advertised gains are not established. read the letter →

arxiv 2504.14783 v2 pith:5GWMAVY7 submitted 2025-04-21 cs.CV cs.AIeess.IVstat.ML

Wenhui Zhu , Peijie Qiu , Xiwen Chen , Zhangsihao Yang , Aristeidis Sotiras , Abolfazl Razi , Yalin Wang This is my paper · ORCID
classification cs.CVcs.AIeess.IVstat.ML
keywords multipleinstancelearningdropoutwholeslideimageclassificationweaksupervisiongradientdirectionerrorflatminimaattentionregularization
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

Multiple instance learning (MIL) labels a whole bag of instances from a single bag label, an approach widely used to classify gigapixel whole-slide images in pathology. Because MIL is usually trained in two stages with a fixed feature extractor, the bag features are noisy and the aggregator can overfit to a few strongly discriminative instances. This paper tries to establish that the standard deep-learning remedy, dropout, works in MIL when applied in a particular way: drop the top-k most important instances rather than random or least important ones. The paper observes that this unusual masking reduces the gradient direction error during training, leads to flatter minima, and improves accuracy and robustness across five MIL benchmarks and two WSI datasets, all with no additional learnable parameters and negligible computational overhead.

What carries the argument

MIL-Dropout is a non-parametric instance-selection mask inserted into the shallow feature extractor of an existing MIL pipeline. At layer $l$, each instance's importance is the sigmoid of its average-pooled embedding; the top-K instances are chosen, and then for each of them the G most cosine-similar instances among the rest are added to the drop set. The mask zeroes those rows of the feature map and rescales the survivors by $\gamma = N/(N-K(1+G))$ to stabilize training. The load-bearing object is not the mask itself but the ranking: the mask removes the currently most attended evidence, forcing the aggregator to discover additional positive instances while the bag label stays unchanged. In experiments the mask is placed after each of three added fully-connected layers before the ABMIL, DSMIL, TransMIL, or DTFD-MIL aggregator.

What would settle it

Take the CAMELYON16 setup and train the same three-layer-feature-extractor ABMIL variant under three conditions: no dropout, random instance dropout of the same mask size, and top-k MIL-Dropout. If random-dropping or no-dropping matches MIL-Dropout in accuracy, F1, AUC, and epsilon-sharpness, then the top-k selection rule is not what carries the reported gains.

Watch

Extended reading notes

Core claim

The central discovery is the counterintuitive empirical finding that, within a bag, removing the instances the model currently considers most important is not harmful but beneficial. A positive bag remains positive after masking, so the network must find alternative, less obvious positive evidence; this behaves like data augmentation and prevents the model from locking onto a single group of easy instances. The paper frames this in optimization terms: among the strategies tested (random, bottom-k, top-k, none), only top-k dropping consistently lowers the gradient direction error relative to the whole-dataset gradient and drives the model to flatter local minima, which is associated with better generalization under Gaussian, Poisson, and salt-and-pepper noise. The paper packages the finding as MIL-Dropout, which computes instance importance with a parameter-free average-pooling attention score, selects the top-K instances plus the G neighbors most similar to each, masks them, and rescales the remaining features.

Load-bearing premise

The load-bearing premise is that the accuracy and robustness gains come from dropping the top-k instances themselves, not from the extra three-layer feature extractor that only the augmented models receive, and the paper includes no control run with that extractor but without the dropout mask.

Editorial extensions

If this is right

  • Plugging MIL-Dropout into the simple ABMIL and ABMIL-Gated aggregators lifts average accuracy by roughly 9.7 to 9.8 points across MUSK1, MUSK2, FOX, TIGER, and ELEPHANT, reaching state-of-the-art numbers on those benchmarks.
  • On CAMELYON16 and TCGA-NSCLC, MIL-Dropout adds on average about 2.0 to 2.5 points of accuracy, F1, and AUC across four MIL aggregators with both ImageNet and self-supervised (SimCLR) features.
  • The method improves robustness to test-time noise and sharpens attention-based lesion localization, so the model's high-attention regions agree better with annotated tumor regions.
  • Because MIL-Dropout adds no learnable parameters and only $\mathcal{O}(N \log N + K(N-K)D^{(l)})$ mask computation per layer, it can be applied to any existing attention-based, transformer-based, or distillation-based MIL aggregator at negligible cost.
  • The paper's extra experiments show MIL-Dropout also improves the concordance index on survival prediction for ABMIL, DSMIL, TransMIL, and DTFD-MIL on TCGA-LUAD and TCGA-BRCA.

Reading between the lines

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

  • If the masking effect is causal, MIL-Dropout is effectively a hard-example mining regularizer: it forces the bag-level model to explain the positive label with evidence outside the current maximum-attention cluster, which should help in tasks where positive evidence is spatially scattered or where easy positives are unrepresentative of the whole disease.
  • The drop-top-k rule could be combined with instance-level uncertainty or with a curriculum that anneals K from large to small, turning the method into a schedule rather than a fixed hyperparameter; the paper does not test this.
  • Because the augmented models in the main comparisons add a three-layer feature extractor that baselines do not have, the cleanest test of the dropout hypothesis would be an ablation that adds the same extractor with dropout disabled; the paper reports no such control.
  • The same importance-ranking idea could be ported to other weakly supervised problems such as video anomaly detection or 3D point-cloud classification, which the paper lists as future work rather than testing.
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

3 major / 6 minor

Summary. The paper studies dropout in multiple instance learning. Section 3 reports investigation experiments showing that, within an ABMIL architecture with a shared three-layer feature extractor, dropping the top-k most important instances yields lower gradient-direction error (GDE), flatter minima under epsilon-sharpness, and better noise robustness than DropNeuron, random, or bottom-k strategies. Based on this, Section 4 proposes MIL-Dropout, which computes instance importance by average pooling over embedding dimensions and drops the top-k instances plus G instances most similar to each top-k instance, with a normalization factor gamma. Sections 5-6 evaluate MIL-Dropout by adding a three-layer feature extractor and the dropout to ABMIL, ABMIL-Gated, DSMIL, TransMIL, and DTFD-MIL on five MIL benchmark datasets and two WSI datasets, reporting large gains (e.g., 9.72% average accuracy on benchmarks and 2-4% AUC on WSI). The appendix includes ablations over K and G, attention-map visualization, a failed attempt to integrate into transformer blocks, and extra survival experiments.

Significance. If the reported effect is causal, MIL-Dropout would be a cheap, architecture-agnostic regularizer for MIL, and the paper's investigation of gradient-direction error, sharpness, and noise robustness would be a useful empirical contribution. The investigation experiments in Section 3 and Appendix A are internally well designed: the same three-layer extractor is used for all dropout variants, and the GDE and sharpness measurements support the top-k observation against DropNeuron. Code is released, and the rebuttal appendix adds extra datasets and survival tasks. However, the main evaluation in Tables 1 and 2 conflates MIL-Dropout with an added trainable feature extractor, so the central quantitative claim is not currently established. The hyperparameter selection on the evaluation datasets further inflates the reported gains. With a proper control arm and a validated selection protocol, the paper's contribution could become solid; as written, the evidence is suggestive rather than conclusive.

major comments (3)
  1. [Section 5.4, Appendix B.1.2, B.2.3, Tables 1-2] The +MIL-Dropout models add a three-layer MLP (256-128-64 on benchmarks; 512/256-256/128-128/64 on WSIs) before the aggregator, while baselines consume raw features (166/230/1024/512 dimensions) directly. No "extra MLP without dropout" control is included, so the gains in Table 1 (e.g., 9.72% average accuracy) and Table 2 (e.g., up to 4% AUC) cannot be attributed to the masking mechanism; they may be due to the extra representation capacity. The investigation experiments in Appendix A use the same extractor for both DropNeuron and DropInstance, which supports the internal comparison, but they also lack an augmented no-dropout arm. Please add control experiments that apply the same f_theta without MIL-Dropout, and ideally with plain neuron dropout, for every aggregator and dataset in Tables 1 and 2.
  2. [Section 6.3, Appendix C.4.1] The optimal values of K and G are identified from Figure 5 on CAMELYON16 and TCGA-NSCLC, and Appendix C.4.1 states that all experiments in Table 2 use these optimal values. The paper does not describe a validation-based selection procedure or a separate test set for hyperparameter selection. If Figure 5 is computed on the same test partitions used in Table 2, the reported gains are optimistically selected. Please either use validation-based selection, report results across a grid without test-set selection, or justify that the chosen K and G are not tuned to the test set.
  3. [Section 3.4, Impact Statement, Conclusion] The paper repeatedly claims "theoretical guarantees" for top-k dropping, but Section 3.4 is a heuristic explanation and the sharpness/GDE analyses are empirical measurements with a standard Taylor expansion; no formal theorem or proof is stated. In addition, the claim that MIL-Dropout can be "seamlessly integrated into existing MIL frameworks" is contradicted by Appendix C.2 and C.3, which report that integration into transformer blocks fails to converge (Table 4 shows 52.5 AUC versus 86.5 for TransMIL). Please temper these claims and specify in the abstract and impact statement that the method is intended for shallow feature extractors before MIL aggregation.
minor comments (6)
  1. [Section 5.2, Table 2] The dataset name "CAMELOYON16" is misspelled and should be "CAMELYON16".
  2. [Section 6.2] The sentence "MIL aggregators trained with features extracted using SimCLR outperforms that trained with features extracted by SimCLR" appears to compare SimCLR with itself; the second instance should presumably be "ImageNet."
  3. [Section 4.0.1] The complexity statement "The complexity is substantially reduced to the fact that O(ND(l)) due to K is typically much smaller than N" is garbled; the stated complexity includes O(K(N-K)D(l)), so with K constant it is O(ND(l)). Please rewrite for clarity.
  4. [Related Work] The sentence "this paper also serves as a theoretical supplement to PDL" is unclear because PDL is not described and no supplement is provided in this manuscript; please either remove the claim or clarify the relationship.
  5. [Appendix C.4.3] The observation that restricting dropout to at most 10% of the average instance count yields consistent improvements is stated only in the rebuttal appendix; if this is a design principle, it should be incorporated into the main text.
  6. [Section 4, Eq. (7)] It is unclear why the sigmoid is applied to the average-pooled values and how the importance scores are normalized across instances before top-k selection; please clarify.

Circularity Check

1 steps flagged · score 2.0 of 10

No derivation-level circularity in the method or its equations; the headline gains are mildly compromised by test-set selection of K and G, which makes them selected rather than predicted.

  1. other [Section 6.3 (ablation) and Appendix C.4.1, applied to Table 2.]
    "All experiments in Table 2 use the optimal values of K and G identified in Figure 5(c)."

    The two hyperparameters K and G are chosen by maximizing AUC on the same CAMELYON16 and TCGA-NSCLC datasets whose accuracy, F1, and AUC gains are later reported in Table 2 as evidence that MIL-Dropout boosts existing MIL methods. The reported gains are therefore the output of a selection procedure over the evaluation data rather than an independent prediction of the method with fixed hyperparameters. This is a post-hoc tuning problem rather than a definitional Eq-X = Eq-Y identity, so it is a mild circularity that lowers the evidential value of the headline claim without collapsing the whole derivation.

full rationale

The paper is an empirical study, and the derivation chain from the top-k DropInstance observation to the MIL-Dropout algorithm is not definitionally circular. Equations (7)-(9) define a new non-parametric importance scorer and a similarity-based query selection; the top-k observation in the investigation experiments uses ABMIL attention maps from Eq. (11), while MIL-Dropout uses a different averaging-based importance, so the method's success is not identical to the input observation. No load-bearing self-citation chain is present: the note that the paper 'serves as a theoretical supplement to PDL' is contextual, and the sharpness/GDE analyses rely on standard external results. The main experimental weakness is a confound: augmented models receive an additional three-layer MLP (Appendix B.1.2 and B.2.3) that baselines do not have, so the causal role of the dropout mask itself is not isolated; this is a correctness/validity concern, not a definitional circularity, and it does not raise the circularity score. Because the only circularity-adjacent issue is the test-set selection of K and G, the score is 2 rather than 0.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The method introduces two hyperparameters (K and G) that must be tuned per dataset; the importance scoring (average pooling) and similarity expansion are heuristics with no independent validation; the paper also depends on standard MIL and sharpness-generalization assumptions.

free parameters (2)
  • K (top-k number) = K = 10 on CAMELYON16, K = 30 on TCGA-NSCLC (Figure 5)
    Number of top-importance instances dropped during training; tuned by ablation per dataset.
  • G (number of similar instances per top-k instance) = Not stated numerically; selected from ablation in Figure 5(b)
    Controls how many similar instances are additionally dropped; tuned per dataset.
assumptions (5)
  • domain assumption MIL weak-supervision model: bag label is positive iff at least one instance is positive (Eq. 1).
    Standard MIL assumption used throughout the paper.
  • standard math Flatter minima generalize better, measured by epsilon-sharpness (Keskar et al., Dinh et al.).
    Invoked in Section 3.2 to interpret sharpness results; no proof re-derived.
  • ad hoc to paper Average pooling over embedding dimensions yields a valid instance importance score (Eq. 7).
    Proposed heuristic, not justified beyond analogy to CBAM; no verification that it matches ABMIL attention ranking.
  • ad hoc to paper Cosine similarity identifies redundant instances whose removal encourages diverse representations (Eq. 8).
    Heuristic; no analysis of sensitivity to this choice.
  • domain assumption The whole-dataset gradient with DropInstance in inference mode is the correct ground truth for GDE.
    Underpins the GDE experiment in Appendix A.1.1; if inference-mode masks are not representative of training, GDE comparisons may mislead.

how reviews work

0 comments
Cite this review

Pith. "Pith review of How Effective Can Dropout Be in Multiple Instance Learning ?." pith.science (2026). https://pith.science/paper/5GWMAVY7

@misc{pith2026250414783,
  author       = {Pith},
  title        = {Pith review of: How Effective Can Dropout Be in Multiple Instance Learning ?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5GWMAVY7}},
  note         = {Machine review of arXiv:2504.14783}
}
read the original abstract

Multiple Instance Learning (MIL) is a popular weakly-supervised method for various applications, with a particular interest in histological whole slide image (WSI) classification. Due to the gigapixel resolution of WSI, applications of MIL in WSI typically necessitate a two-stage training scheme: first, extract features from the pre-trained backbone and then perform MIL aggregation. However, it is well-known that this suboptimal training scheme suffers from "noisy" feature embeddings from the backbone and inherent weak supervision, hindering MIL from learning rich and generalizable features. However, the most commonly used technique (i.e., dropout) for mitigating this issue has yet to be explored in MIL. In this paper, we empirically explore how effective the dropout can be in MIL. Interestingly, we observe that dropping the top-k most important instances within a bag leads to better performance and generalization even under noise attack. Based on this key observation, we propose a novel MIL-specific dropout method, termed MIL-Dropout, which systematically determines which instances to drop. Experiments on five MIL benchmark datasets and two WSI datasets demonstrate that MIL-Dropout boosts the performance of current MIL methods with a negligible computational cost. The code is available at https://github.com/ChongQingNoSubway/MILDropout.

Figures

Figures reproduced from arXiv: 2504.14783 by the authors.

Figure 1
Figure 1. An illustrative example comparing the convergence tra￾jectories of the baseline ABMIL without dropout (Left) and AB￾MIL with the proposed MIL-Dropout (Right). ABMIL without dropout is likely to follow an incorrect gradient direction initially and eventually converge to a sharp minimum. In contrast, ABMIL with the proposed MIL-Dropout typically achieves a lower gra￾dient direction error and reaches a flatter minimum … view at source ↗
Figure 2
Figure 2. An conceptual illustration for a flat and sharp minimum in 1D curvature (Left) and 2D landscape (Right) of the loss func￾tion Lθ. to its popularity. The standard workflow of an embedding￾based MIL involves (i) projecting instances into feature em￾beddings via an instance-level feature extractor and (ii) ag￾gregating the instance-level features into a bag-level predic￾tion through an MIL aggregator. Specifically, the… view at source ↗
Figure 3
Figure 3. The landscape of the loss function Lθ for two different dropout strategies (a) DropNeuron and (b) DropInstance as well as (c) the performance of MIL models against different noise attacks. We mark the Euclidean ball B2(ϵ, θ∗ ) around the optimal parameter θ ∗ (see Eq. 5) in subpanel figure (a) and (b) with a red circle. We observe that the landscape of Lθ in the DropInstance scenario leads to flatter minima compared… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: The comparison of change of GDE (Left) over the first 10,000 iterations as well as performance and loss (line plot) and AUC (bar plot) when using different instance dropout strategies (Right), where the area under GDE is the area enclosed by GDE and the x-axis. Droppin…
Figure 5
Figure 5. Figure 5: Ablation studies on the number of top-k instances K (a) and similarity instance S (b) using CAMELYON16 and TCGA-NSCLC datasets. (c) Attention map from ABMIL without and with MIL-Dropout, with tumor regions outlined in red. Brighter cyan in columns two and three indicat…
Figure 6
Figure 6. Figure 6: ABMIL aggregation with Dropout setting within investigation experiments . A. Investigation Experiment Design We implementation all experiments on a node of cluster with NVIDIA V100 (32GB). We use Pytorch Library (Paszke et al., 2019) with version of 1.13. The investiga…
Figure 7
Figure 7. Figure 7: The exemplary patch with different noise types and strengths [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: Different aggregation methods with MIL-Dropout setting within main experiments . obtained from the previous iteration: ρψ({vn | n = 1, · · · , N}) = X N n=1 αnvn, with αn = softmax(wT 1 tanh(w2v T n )), (11) where αn implies the importance of the n-th instance. In this…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 27 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    Support vector machines for multiple-instance learning

    Andrews, S., Tsochantaridis, I., and Hofmann, T. Support vector machines for multiple-instance learning. Advances in neural information processing systems, 15, 2002

    work page 2002

  3. [3]

    Robust object tracking with online multiple instance learning

    Babenko, B., Yang, M.-H., and Belongie, S. Robust object tracking with online multiple instance learning. IEEE transactions on pattern analysis and machine intelligence, 33 0 (8): 0 1619--1632, 2010

    work page 2010

  4. [4]

    J., Ding, T., Lu, M

    Chen, R. J., Ding, T., Lu, M. Y., Williamson, D. F., Jaume, G., Chen, B., Zhang, A., Shao, D., Song, A. H., Shaban, M., et al. Towards a general-purpose foundation model for computational pathology. Nature Medicine, 2024 a

    work page 2024

  5. [5]

    A simple framework for contrastive learning of visual representations

    Chen, T., Kornblith, S., Norouzi, M., and Hinton, G. A simple framework for contrastive learning of visual representations. In International conference on machine learning, pp.\ 1597--1607. PMLR, 2020

    2020

  6. [6]

    Timemil: advancing multivariate time series classification via a time-aware multiple instance learning

    Chen, X., Qiu, P., Zhu, W., Li, H., Wang, H., Sotiras, A., Wang, Y., and Razi, A. Timemil: advancing multivariate time series classification via a time-aware multiple instance learning. In Proceedings of the 41st International Conference on Machine Learning, pp.\ 7190--7206, 2024 b

    work page 2024

  7. [7]

    and Shim, H

    Choe, J. and Shim, H. Attention-based dropout layer for weakly supervised object localization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 2219--2228, 2019

    work page 2019

  8. [8]

    G., Lathrop, R

    Dietterich, T. G., Lathrop, R. H., and Lozano-P \'e rez, T. Solving the multiple instance problem with axis-parallel rectangles. Artificial intelligence, 89 0 (1-2): 0 31--71, 1997

    work page 1997

Show all 47 references
  1. [9]

    Sharp minima can generalize for deep nets

    Dinh, L., Pascanu, R., Bengio, S., and Bengio, Y. Sharp minima can generalize for deep nets. In International Conference on Machine Learning, pp.\ 1019--1028. PMLR, 2017

    2017

  2. [10]

    K., Cutajar, K., Xie, H., Kandola, J., and Twomey, N

    Early, J., Cheung, G. K., Cutajar, K., Xie, H., Kandola, J., and Twomey, N. Inherently interpretable time series classification via multiple instance learning. arXiv preprint arXiv:2311.10049, 2023

    2023 arXiv

  3. [11]

    and Zhou, Z.-H

    Feng, J. and Zhou, Z.-H. Deep miml network. In Proceedings of the AAAI conference on artificial intelligence, volume 31, 2017

    2017

  4. [12]

    Sharpness-aware minimization for efficiently improving generalization

    Foret, P., Kleiner, A., Mobahi, H., and Neyshabur, B. Sharpness-aware minimization for efficiently improving generalization. In International Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=6Tm1mposlrM

    2021

  5. [13]

    Ghiasi, G., Lin, T.-Y., and Le, Q. V. Dropblock: A regularization method for convolutional networks. Advances in neural information processing systems, 31, 2018

    2018

  6. [14]

    E., Srivastava, N., Krizhevsky, A., Sutskever, I., and Salakhutdinov, R

    Hinton, G. E., Srivastava, N., Krizhevsky, A., Sutskever, I., and Salakhutdinov, R. R. Improving neural networks by preventing co-adaptation of feature detectors. arXiv preprint arXiv:1207.0580, 2012

    2012 arXiv

  7. [15]

    M., Gao, Y., Davis, J

    Hou, L., Samaras, D., Kurc, T. M., Gao, Y., Davis, J. E., and Saltz, J. H. Patch-based convolutional neural network for whole slide tissue image classification. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp.\ 2424--2433, 2016

    2016

  8. [16]

    Unleash the power of state space model for whole slide image with local aware scanning and importance resampling

    Huang, Y., Zhao, W., Fu, Y., Zhu, L., and Yu, L. Unleash the power of state space model for whole slide image with local aware scanning and importance resampling. IEEE Transactions on Medical Imaging, 2024

    2024

  9. [17]

    Attention-based deep multiple instance learning

    Ilse, M., Tomczak, J., and Welling, M. Attention-based deep multiple instance learning. In International conference on machine learning, pp.\ 2127--2136. PMLR, 2018

    2018

  10. [18]

    The break-even point on optimization trajectories of deep neural networks

    Jastrzebski, S., Szymczak, M., Fort, S., Arpit, D., Tabor, J., Cho, K., and Geras, K. The break-even point on optimization trajectories of deep neural networks. arXiv preprint arXiv:2002.09572, 2020

    2002 arXiv

  11. [19]

    J., Williamson, D

    Jaume, G., Vaidya, A., Chen, R. J., Williamson, D. F., Liang, P. P., and Mahmood, F. Modeling dense multimodal interactions between biological pathways and histology for survival prediction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, ...

    2024

  12. [20]

    T., and Cevher, V

    Kavis, A., Skoulakis, S., Antonakopoulos, K., Dadi, L. T., and Cevher, V. Adaptive stochastic variance reduction for non-convex finite-sum minimization. Advances in Neural Information Processing Systems, 35: 0 23524--23538, 2022

    2022

  13. [21]

    S., Mudigere, D., Nocedal, J., Smelyanskiy, M., and Tang, P

    Keskar, N. S., Mudigere, D., Nocedal, J., Smelyanskiy, M., and Tang, P. T. P. On large-batch training for deep learning: Generalization gap and sharp minima. arXiv preprint arXiv:1609.04836, 2016

    2016 arXiv

  14. [22]

    Li, B., Li, Y., and Eliceiri, K. W. Dual-stream multiple instance learning network for whole slide image classification with self-supervised contrastive learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 14318--14328, 2021

    2021

  15. [23]

    Dropout reduces underfitting

    Liu, Z., Xu, Z., Jin, J., Shen, Z., and Darrell, T. Dropout reduces underfitting. In International Conference on Machine Learning, pp.\ 22233--22248. PMLR, 2023

    2023

  16. [24]

    Park, J., Woo, S., Lee, J.-Y., and Kweon, I. S. Bam: Bottleneck attention module. arXiv preprint arXiv:1807.06514, 2018

    2018 arXiv

  17. [25]

    Pytorch: An imperative style, high-performance deep learning library

    Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., et al. Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing systems, 32, 2019

    2019

  18. [26]

    Sc-mil: Sparsely coded multiple instance learning for whole slide image classification

    Qiu, P., Xiao, P., Zhu, W., Wang, Y., and Sotiras, A. Sc-mil: Sparsely coded multiple instance learning for whole slide image classification. arXiv preprint arXiv:2311.00048, 2023

    2023 arXiv

  19. [27]

    Boosting whole slide image classification from the perspectives of distribution, correlation and magnification

    Qu, L., Yang, Z., Duan, M., Ma, Y., Wang, S., Wang, M., and Song, Z. Boosting whole slide image classification from the perspectives of distribution, correlation and magnification. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp.\ 21463--21473, 2023

    2023

  20. [28]

    Multiple-instance learning for medical image and video analysis

    Quellec, G., Cazuguel, G., Cochener, B., and Lamard, M. Multiple-instance learning for medical image and video analysis. IEEE reviews in biomedical engineering, 10: 0 213--234, 2017

    2017

  21. [29]

    Transmil: Transformer based correlated multiple instance learning for whole slide image classification

    Shao, Z., Bian, H., Chen, Y., Wang, Y., Zhang, J., Ji, X., et al. Transmil: Transformer based correlated multiple instance learning for whole slide image classification. Advances in neural information processing systems, 34: 0 2136--2147, 2021

    2021

  22. [30]

    H., Chen, R

    Song, A. H., Chen, R. J., Jaume, G., Vaidya, A. J., Baras, A. S., and Mahmood, F. Multimodal prototyping for cancer survival prediction. arXiv preprint arXiv:2407.00224, 2024

    2024 arXiv

  23. [31]

    Dropout: A simple way to prevent neural networks from overfitting

    Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., and Salakhutdinov, R. Dropout: A simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15 0 (56): 0 1929--1958, 2014. URL http://jmlr.org/papers/v15/srivastava14a.html

    1929

  24. [32]

    Multiple instance learning framework with masked hard instance mining for whole slide image classification

    Tang, W., Huang, S., Zhang, X., Zhou, F., Zhang, Y., and Liu, B. Multiple instance learning framework with masked hard instance mining for whole slide image classification. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp.\ 4078--4087, 2023

    2023

  25. [33]

    Efficient object localization using convolutional networks

    Tompson, J., Goroshin, R., Jain, A., LeCun, Y., and Bregler, C. Efficient object localization using convolutional networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp.\ 648--656, 2015

    2015

  26. [35]

    Multiple instance learning with graph neural networks

    Tu, M., Huang, J., He, X., and Zhou, B. Multiple instance learning with graph neural networks. arXiv preprint arXiv:1906.04881, 2019 b

    1906 arXiv

  27. [36]

    Non-local neural networks

    Wang, X., Girshick, R., Gupta, A., and He, K. Non-local neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp.\ 7794--7803, 2018 a

    2018

  28. [37]

    Revisiting multiple instance neural networks

    Wang, X., Yan, Y., Tang, P., Bai, X., and Liu, W. Revisiting multiple instance neural networks. Pattern Recognition, 74: 0 15--24, 2018 b

    2018

  29. [38]

    Woo, S., Park, J., Lee, J.-Y., and Kweon, I. S. Cbam: Convolutional block attention module. In Proceedings of the European conference on computer vision (ECCV), pp.\ 3--19, 2018

    2018

  30. [39]

    and Zhang, J

    Xiang, J. and Zhang, J. Exploring low-rank property in multiple instance learning for whole slide image classification. In The Eleventh International Conference on Learning Representations, 2023

    2023

  31. [40]

    Camel: A weakly supervised learning framework for histopathology image segmentation

    Xu, G., Song, Z., Sun, Z., Ku, C., Yang, Z., Liu, C., Wang, S., Ma, J., and Xu, W. Camel: A weakly supervised learning framework for histopathology image segmentation. In Proceedings of the IEEE/CVF International Conference on computer vision, pp.\ 10682--10691, 2019

    2019

  32. [41]

    Deep multi-instance learning with dynamic pooling

    Yan, Y., Wang, X., Guo, X., Fang, J., Liu, W., and Huang, J. Deep multi-instance learning with dynamic pooling. In Asian Conference on Machine Learning, pp.\ 662--677. PMLR, 2018

    2018

  33. [42]

    E., and Zheng, Y

    Zhang, H., Meng, Y., Zhao, Y., Qiao, Y., Yang, X., Coupland, S. E., and Zheng, Y. Dtfd-mil: Double-tier feature distillation multiple instance learning for histopathology whole slide image classification. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern...

    2022

  34. [43]

    and Xu, Z.-Q

    Zhang, Z. and Xu, Z.-Q. J. Implicit regularization of dropout. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    2024

  35. [44]

    Dynamic policy-driven adaptive multi-instance learning for whole slide image classification

    Zheng, T., Jiang, K., and Yao, H. Dynamic policy-driven adaptive multi-instance learning for whole slide image classification. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 8028--8037, 2024

    2024

  36. [45]

    M., and Wang, Y

    Zhu, W., Qiu, P., Chen, X., Dumitrascu, O. M., and Wang, Y. Pdl: Regularizing multiple instance learning with progressive dropout layers. arXiv preprint arXiv:2308.10112, 2023 a

    2023 arXiv

  37. [46]

    M., and Wang, Y

    Zhu, W., Qiu, P., Lepore, N., Dumitrascu, O. M., and Wang, Y. Self-supervised equivariant regularization reconciles multiple-instance learning: Joint referable diabetic retinopathy classification and lesion segmentation. In 18th International Symposium on Medical Information P...

    2023

  38. [47]

    Dgr-mil: Exploring diverse global representation in multiple instance learning for whole slide image classification

    Zhu, W., Chen, X., Qiu, P., Sotiras, A., Razi, A., and Wang, Y. Dgr-mil: Exploring diverse global representation in multiple instance learning for whole slide image classification. In European Conference on Computer Vision, pp.\ 333--351. Springer, 2024

    2024

  39. [48]

    Asymmetric non-local neural networks for semantic segmentation

    Zhu, Z., Xu, M., Bai, S., Huang, T., and Bai, X. Asymmetric non-local neural networks for semantic segmentation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp.\ 593--602, 2019

    2019

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.