Pith. sign in

REVIEW 2 major objections 2 minor 51 references

FraPPE: Fast and Efficient Preference-based Pure Exploration

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

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

Pith's one-line read A new bandit algorithm identifies Pareto-optimal arms sample-optimally for any preference cone, with O(KL²) computation time.

desk verdict The abstract promises a genuinely useful bandit result, but the supplied full text is an unrelated histology paper, so there is no manuscript to review. read the letter →

arxiv 2508.16487 v1 pith:LEIFOSOG submitted 2025-08-22 cs.LG cs.AImath.OCmath.STstat.MLstat.TH

Udvas Das , Apurv Shukla , Debabrota Basu This is my paper · ORCID
classification cs.LGcs.AImath.OCmath.STstat.MLstat.TH MSC 62L0590C2990C52
keywords preference-basedpureexplorationmulti-objectivebanditsParetooptimalsetpreferenceconesamplecomplexitylowerboundFrank-Wolfeoptimizationmaxmin
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

Preference-based Pure Exploration (PrePEx) asks which arms of a multi-objective bandit are Pareto-optimal under a given preference cone. This paper proposes FraPPE, an algorithm that asymptotically matches the known lower bound on the number of samples needed to answer that question with confidence, for any arbitrary preference cone. The central technical move is to make the lower bound's maxmin optimization tractable: three structural properties reduce the minimization side, and a Frank-Wolfe optimizer accelerates the maximization side. Together they solve the maxmin problem in O(KL²) time for K arms and L reward dimensions, a speedup over earlier algorithms. If FraPPE is correct, sample-optimal Pareto-set identification is no longer computationally out of reach.

What carries the argument

The central object is the maxmin optimization inside the PrePEx lower bound: minimize over plausible Pareto-set configurations and maximize over reward distributions consistent with the instance. The machinery is a two-part attack on this problem. Three structural properties of the lower bound turn the minimization into a computationally tractable reduction, while a Frank-Wolfe optimizer, a conditional-gradient method suited to convex and smooth objectives with sparse iterates, accelerates the maximization. The combination is what brings the per-iteration cost to O(KL²) and lets the algorithm track the lower bound asymptotically.

What would settle it

Run FraPPE on a preference cone and arm reward distributions where the lower-bound maximization is known to be non-convex; if the optimizer stops at a local optimum and the sample complexity exceeds the lower bound, the claim fails. Concretely, a synthetic instance with K=3, L=2 and a pointed cone that creates a saddle in the maxmin objective is a direct test of the O(KL²) and asymptotic optimality claims.

Watch

Extended reading notes

Core claim

FraPPE, a PrePEx algorithm, is the first computationally efficient algorithm that achieves asymptotic sample optimality for arbitrary preference cones. It tracks the existing lower bound by solving the lower bound's maxmin problem in O(KL²) time: the minimization over candidate Pareto sets is reduced to a tractable form using three structural properties of the lower bound, and the maximization over reward distributions is handled by a Frank-Wolfe optimizer. The paper proves that FraPPE asymptotically achieves the optimal sample complexity, and experiments on synthetic and real datasets report the lowest sample complexities among existing PrePEx algorithms for exact Pareto-set identification.

Load-bearing premise

The runtime and sample-optimality guarantees both depend on the Frank-Wolfe optimizer reliably reaching the global maximum of the lower-bound problem; if the optimizer can get stuck at a local maximum, the speed and the sample guarantee do not follow.

Editorial extensions

If this is right

  • Sample-optimal Pareto-set identification becomes feasible for bandits with many arms and reward dimensions, because the per-decision cost no longer grows intractably.
  • Arbitrary preference cones, not just the positive orthant, can be handled without a computational penalty, widening applicability to lexicographic, polyhedral, or other dominance orders.
  • The O(KL²) maxmin solver can be reused as a subroutine in other pure-exploration algorithms whose lower bounds share the same maxmin structure.
  • FraPPE's sample complexity matches the lower bound asymptotically, so further gains must come from constant factors or a different problem formulation.

Reading between the lines

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

  • Frank-Wolfe's convergence depends on the lower-bound maximization objective being convex and smooth; the abstract does not state those regularity conditions, so the O(KL²) guarantee and the asymptotic optimality are conditional on them.
  • The three structural properties of the lower bound may transfer to other active-learning or ranking problems whose objective can be written as a maxmin over a structured family, enabling similar speedups.
  • A testable extension is to benchmark FraPPE against the lower bound on synthetic instances with a non-polyhedral cone, such as a Lorentz cone, where the Frank-Wolfe acceleration may require more iterations or stall at a local optimum.
  • If Frank-Wolfe's iterates are sparse, FraPPE might effectively identify not only the Pareto set but also a small set of arms and reward directions that dominate the sample cost, an interpretability byproduct the authors do not advertise.
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

2 major / 2 minor

Summary. The submission is indexed as arXiv:2508.16487 (cs.LG), titled 'FraPPE: Fast and Efficient Preference-based Pure Exploration.' The abstract claims a new algorithm, FraPPE, for preference-based pure exploration in vector-valued bandits, with three structural properties of a lower bound enabling a tractable minimisation reduction, a Frank-Wolfe optimiser accelerating the maximisation, an overall O(KL^2) time for solving the maxmin problem, and asymptotic optimal sample complexity matching an existing lower bound for arbitrary preference cones. However, the supplied full text is not this manuscript. It is an IEEE Transactions on Medical Imaging paper titled 'Disentangled Multi-modal Learning of Histology and Transcriptomics for Cancer Characterization' (arXiv:2508.16479v2), with a different abstract, methods, experiments, and references. Consequently, none of the claimed algorithmic contributions, proofs, runtime analyses, or experimental comparisons for FraPPE are present in the submitted text.

Significance. If the claims in the abstract were substantiated, the work would address a genuine open problem: no existing PrePEx algorithm is both sample-optimal and computationally efficient for arbitrary preference cones. The claimed O(KL^2) maxmin solver and asymptotic sample optimality would be a notable advance for multi-objective pure exploration. However, the submitted manuscript body contains no algorithm statement, no theoretical derivation, no proofs, and no experiments for FraPPE. The central claims are therefore entirely unsupported by the supplied text, and their significance cannot be assessed. This is not a case of a defensible claim with local gaps; the evidence base for every load-bearing assertion is absent.

major comments (2)
  1. [Full Text (entire body)] The supplied full text is an unrelated medical-imaging paper: 'Disentangled Multi-modal Learning of Histology and Transcriptomics for Cancer Characterization,' with footer identifier arXiv:2508.16479v2, while the submission is arXiv:2508.16487. It contains no definition of FraPPE, no preference-cone bandit setup, no lower-bound equations, no Frank-Wolfe analysis, and no sample-complexity theorem. The abstract's central claims—asymptotic optimal sample complexity and O(KL^2) maxmin solving—are thus unsupported by any derivable evidence in the manuscript. This is a load-bearing defect that invalidates the submission as a reviewer artifact.
  2. [Abstract] The abstract asserts 'three structural properties of the lower bound' and a Frank-Wolfe-based maximisation acceleration, but none of these properties, the resulting reduction, or the convergence conditions for Frank-Wolfe are presented anywhere in the submitted text. In particular, the claimed O(KL^2) runtime depends on the unstated convexity/smoothness and global-optimality conditions of the maxmin problem. Since the body is a different paper, not even the notation K, L, and C is defined. These are not presentation issues; they are the core technical content needed to verify the claims.
minor comments (2)
  1. [Full Text footer] The manuscript footer shows 'arXiv:2508.16479v2 [eess.IV] 28 Feb 2026,' inconsistent with the stated submission ID 2508.16487. This mismatch corroborates that the wrong full text was provided.
  2. [Title/Abstract] The title and abstract refer to FraPPE and preference-based pure exploration, whereas the body is titled and written about histology/transcriptomics multi-modal learning. The author names on the supplied body also do not match those implied by the FraPPE submission.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity detectable: supplied full text is an unrelated medical-imaging manuscript, so the FraPPE derivation is not present to analyze.

full rationale

The abstract describes FraPPE's claimed O(KL^2) maxmin solver and asymptotic optimality, but the supplied full text is an IEEE TMI paper, "Disentangled Multi-modal Learning of Histology and Transcriptomics for Cancer Characterization," with different authors and arXiv identifier 2508.16479v2. None of the FraPPE paper's own equations are available: the three structural properties of the lower bound, the Frank-Wolfe optimization details, the sample-complexity proof, and the experiments are all absent. Under the hard rule that circularity must be exhibited by quoting the specific reduction (e.g., Eq. X = Eq. Y by construction, or a fitted parameter renamed as a prediction), no such reduction can be identified from the evidence provided. The mismatch is a provenance/correctness concern, not a circularity finding. Therefore the score is 0 with no steps listed; the manuscript cannot be assessed for circularity because its actual derivation chain is not present in the supplied text.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

No free parameters are identifiable from the abstract alone: K arms, L reward dimensions, and the preference cone C are problem inputs, not fitted values. FraPPE is an algorithm, not a postulated entity, so no physical or mathematical objects are invented by the abstract. The three axioms listed are the unstated premises the abstract's claims rest on. A full audit requires the actual manuscript, including its algorithm statement, proofs, and experiments, none of which appear in the supplied text.

assumptions (3)
  • domain assumption Rewards are vector-valued and ordered by a fixed, given preference cone C, in the standard PrePEx bandit model.
    From the abstract: 'reward vectors are ordered via a (given) preference cone C'. The sample-complexity lower bound and the Pareto-optimality notion both depend on this model.
  • ad hoc to paper The three structural properties of the lower bound stated in the abstract are true and yield an exact, not approximate, reduction of the minimisation problem.
    Abstract: 'we derive three structural properties of the lower bound that yield a computationally tractable reduction'. These properties are asserted without statement and are load-bearing for the O(KL^2) runtime claim.
  • ad hoc to paper The lower-bound maximisation subproblem is convex and smooth enough that a Frank-Wolfe optimiser reaches its global maximum in O(KL^2) total work.
    Abstract: 'we deploy a Frank-Wolfe optimiser to accelerate the maximisation problem'. Global convergence of Frank-Wolfe requires convexity and appropriate step sizes; the abstract states no such conditions, and none of the body text addresses them.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FraPPE: Fast and Efficient Preference-based Pure Exploration." pith.science (2026). https://pith.science/paper/LEIFOSOG

@misc{pith2026250816487,
  author       = {Pith},
  title        = {Pith review of: FraPPE: Fast and Efficient Preference-based Pure Exploration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LEIFOSOG}},
  note         = {Machine review of arXiv:2508.16487}
}
abstract

Preference-based Pure Exploration (PrePEx) aims to identify with a given confidence level the set of Pareto optimal arms in a vector-valued (aka multi-objective) bandit, where the reward vectors are ordered via a (given) preference cone $\mathcal{C}$. Though PrePEx and its variants are well-studied, there does not exist a computationally efficient algorithm that can optimally track the existing lower bound for arbitrary preference cones. We successfully fill this gap by efficiently solving the minimisation and maximisation problems in the lower bound. First, we derive three structural properties of the lower bound that yield a computationally tractable reduction of the minimisation problem. Then, we deploy a Frank-Wolfe optimiser to accelerate the maximisation problem in the lower bound. Together, these techniques solve the maxmin optimisation problem in $\mathcal{O}(KL^{2})$ time for a bandit instance with $K$ arms and $L$ dimensional reward, which is a significant acceleration over the literature. We further prove that our proposed PrePEx algorithm, FraPPE, asymptotically achieves the optimal sample complexity. Finally, we perform numerical experiments across synthetic and real datasets demonstrating that FraPPE achieves the lowest sample complexities to identify the exact Pareto set among the existing algorithms.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 43 canonical work pages

  1. [1]

    Predicting survival from colorectal cancer histology slides using deep learning: A retrospective multicenter study,

    J. N. Katheret al., “Predicting survival from colorectal cancer histology slides using deep learning: A retrospective multicenter study,”PLoS medicine, vol. 16, no. 1, p. e1002730, 2019

    work page 2019

  2. [2]

    Scaling vision transformers to gigapixel images via hierarchical self-supervised learning,

    R. J. Chenet al., “Scaling vision transformers to gigapixel images via hierarchical self-supervised learning,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 16 144–16 155

    work page 2022

  3. [3]

    Attention-based deep multiple instance learning,

    M. Ilse, J. Tomczak, and M. Welling, “Attention-based deep multiple instance learning,” inInternational conference on machine learning. PMLR, 2018, pp. 2127–2136

    2018

  4. [4]

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

    Z. Shaoet al., “Transmil: Transformer based correlated multiple instance learning for whole slide image classification,”Advances in neural information processing systems, vol. 34, pp. 2136–2147, 2021

    work page 2021

  5. [5]

    Dtfd-mil: Double-tier feature distillation multiple instance learning for histopathology whole slide image classification,

    H. Zhanget al., “Dtfd-mil: Double-tier feature distillation multiple instance learning for histopathology whole slide image classification,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 18 802–18 812

    work page 2022

  6. [6]

    Data-efficient and weakly supervised computational pathology on whole-slide images,

    M. Y . Lu, D. F. Williamson, T. Y . Chen, R. J. Chen, M. Barbieri, and F. Mahmood, “Data-efficient and weakly supervised computational pathology on whole-slide images,”Nature biomedical engineering, vol. 5, no. 6, pp. 555–570, 2021

    2021

  7. [7]

    Dynamic graph representation with knowledge-aware attention for histopathology whole slide image analysis,

    J. Liet al., “Dynamic graph representation with knowledge-aware attention for histopathology whole slide image analysis,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2024, pp. 11 323–11 332

    work page 2024

  8. [8]

    Histopathology whole slide image analysis with heterogeneous graph representation learning,

    T. H. Chan, F. J. Cendra, L. Ma, G. Yin, and L. Yu, “Histopathology whole slide image analysis with heterogeneous graph representation learning,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 15 661–15 670

    work page 2023

Show all 51 references
  1. [9]

    A graph-based multi-scale approach with knowledge distillation for wsi classification,

    G. Bontempo, F. Bolelli, A. Porrello, S. Calderara, and E. Ficarra, “A graph-based multi-scale approach with knowledge distillation for wsi classification,”IEEE Transactions on Medical Imaging, vol. 43, no. 4, pp. 1412–1421, 2023

    2023

  2. [10]

    Multimodal cancer survival analysis via hypergraph learn- ing with cross-modality rebalance,

    M. Quet al., “Multimodal cancer survival analysis via hypergraph learn- ing with cross-modality rebalance,”arXiv preprint arXiv:2505.11997, 2025

    2025 arXiv

  3. [11]

    Multimodal co-attention transformer for survival prediction in gigapixel whole slide images,

    R. J. Chenet al., “Multimodal co-attention transformer for survival prediction in gigapixel whole slide images,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 4015–4025

    2021

  4. [12]

    Multimodal optimal transport-based co-attention transformer with global structure consistency for survival prediction,

    Y . Xu and H. Chen, “Multimodal optimal transport-based co-attention transformer with global structure consistency for survival prediction,” inProceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 21 241–21 251

    2023

  5. [13]

    Memory-augmented incomplete multimodal survival prediction via cross-slide and gene-attentive hypergraph learning,

    M. Quet al., “Memory-augmented incomplete multimodal survival prediction via cross-slide and gene-attentive hypergraph learning,” in International Conference on Medical Image Computing and Computer- Assisted Intervention. Springer, 2025, pp. 318–327

    2025

  6. [14]

    Modeling dense multimodal interactions between biolog- ical pathways and histology for survival prediction,

    G. Jaume, A. Vaidya, R. J. Chen, D. F. Williamson, P. P. Liang, and F. Mahmood, “Modeling dense multimodal interactions between biolog- ical pathways and histology for survival prediction,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 20...

    2024

  7. [15]

    Prototypical infor- mation bottlenecking and disentangling for multimodal cancer survival prediction,

    Y . Zhang, Y . Xu, J. Chen, F. Xie, and H. Chen, “Prototypical infor- mation bottlenecking and disentangling for multimodal cancer survival prediction,” inICLR, 2024

    2024

  8. [16]

    Multimodal prototyping for cancer survival prediction,

    A. H. Song, R. J. Chen, G. Jaume, A. J. Vaidya, A. Baras, and F. Mahmood, “Multimodal prototyping for cancer survival prediction,” inInternational Conference on Machine Learning. PMLR, 2024, pp. 46 050–46 073

    2024

  9. [17]

    Cross-modal translation and alignment for survival analysis,

    F. Zhou and H. Chen, “Cross-modal translation and alignment for survival analysis,” inProceedings of the IEEE/CVF International Con- ference on Computer Vision, 2023, pp. 21 485–21 494

    2023

  10. [18]

    Mome: Mixture of multimodal experts for cancer survival prediction,

    C. Xionget al., “Mome: Mixture of multimodal experts for cancer survival prediction,” inInternational Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2024, pp. 318–328

    2024

  11. [19]

    Deciphering tumor ecosystems at super resolution from spatial transcriptomics with tesla,

    J. Huet al., “Deciphering tumor ecosystems at super resolution from spatial transcriptomics with tesla,”Cell systems, vol. 14, no. 5, pp. 404– 417, 2023

    2023

  12. [20]

    Pathomic fusion: an integrated framework for fusing histopathology and genomic features for cancer diagnosis and prognosis,

    R. J. Chenet al., “Pathomic fusion: an integrated framework for fusing histopathology and genomic features for cancer diagnosis and prognosis,”IEEE Transactions on Medical Imaging, vol. 41, no. 4, pp. 757–770, 2020

    2020

  13. [21]

    Discrepancy and gradient-guided multi-modal knowledge distillation for patholog- ical glioma grading,

    X. Xing, Z. Chen, M. Zhu, Y . Hou, Z. Gao, and Y . Yuan, “Discrepancy and gradient-guided multi-modal knowledge distillation for patholog- ical glioma grading,” inInternational Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2022, pp. 636–646

    2022

  14. [22]

    Deep learning model with pathological knowledge for detection of colorectal neuroendocrine tumor,

    K. Zhenget al., “Deep learning model with pathological knowledge for detection of colorectal neuroendocrine tumor,”Cell Reports Medicine, vol. 5, no. 10, 2024

    2024

  15. [23]

    Distribution based mil pooling filters: Experiments on a lymph node metastases dataset,

    M. U. Oner, J. M. S. Kye-Jet, H. K. Lee, and W.-K. Sung, “Distribution based mil pooling filters: Experiments on a lymph node metastases dataset,”Medical Image Analysis, vol. 87, p. 102813, 2023

    2023

  16. [24]

    Clinical-grade computational pathology us- ing weakly supervised deep learning on whole slide images,

    G. Campanellaet al., “Clinical-grade computational pathology us- ing weakly supervised deep learning on whole slide images,”Nature medicine, vol. 25, no. 8, pp. 1301–1309, 2019

    2019

  17. [25]

    Knowledge-driven subspace fusion and gradient coordination for multi-modal learning,

    Y . Zhang, X. Wang, F. Meng, J. Tang, and C. Li, “Knowledge-driven subspace fusion and gradient coordination for multi-modal learning,” in International Conference on Medical Image Computing and Computer- Assisted Intervention. Springer, 2024, pp. 263–273

    2024

  18. [26]

    Focus on focus: Focus-oriented representation learn- ing and multi-view cross-modal alignment for glioma grading,

    L. Panet al., “Focus on focus: Focus-oriented representation learn- ing and multi-view cross-modal alignment for glioma grading,”arXiv preprint arXiv:2408.08527, 2024

    2024 arXiv

  19. [27]

    Classification and mutation prediction from non– small cell lung cancer histopathology images using deep learning,

    N. Coudrayet al., “Classification and mutation prediction from non– small cell lung cancer histopathology images using deep learning,” Nature medicine, vol. 24, no. 10, pp. 1559–1567, 2018

    2018

  20. [28]

    Beyond genetics: driving cancer with the tumour microenvironment behind the wheel,

    S. Yuan, J. Almagro, and E. Fuchs, “Beyond genetics: driving cancer with the tumour microenvironment behind the wheel,”Nature Reviews Cancer, vol. 24, no. 4, pp. 274–286, 2024

    2024

  21. [29]

    Multi-task learning of histology and molecular markers for classifying diffuse glioma,

    X. Wang, S. Price, and C. Li, “Multi-task learning of histology and molecular markers for classifying diffuse glioma,” inInternational Conference on Medical Image Computing and Computer-Assisted In- tervention. Springer, 2023, pp. 551–561

    2023

  22. [30]

    Dual-space disentangled- multimodal network (ddm-net) for glioma diagnosis and prognosis with incomplete pathology and genomic data,

    L. Qiu, L. Zhao, W. Zhao, and J. Zhao, “Dual-space disentangled- multimodal network (ddm-net) for glioma diagnosis and prognosis with incomplete pathology and genomic data,”Physics in Medicine & Biology, vol. 69, no. 8, p. 085028, 2024

    2024

  23. [31]

    Robust multimodal survival prediction with conditional latent differentiation variational autoencoder,

    J. Zhou, J. Tang, Y . Zuo, P. Wan, D. Zhang, and W. Shao, “Robust multimodal survival prediction with conditional latent differentiation variational autoencoder,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 10 384–10 393

    2025

  24. [32]

    Histo- genomic knowledge association for cancer prognosis from histopathol- ogy whole slide images,

    Z. Wang, Y . Zhang, Y . Xu, S. Imoto, H. Chen, and J. Song, “Histo- genomic knowledge association for cancer prognosis from histopathol- ogy whole slide images,”IEEE Transactions on Medical Imaging, 2025

    2025

  25. [33]

    Self- normalizing neural networks,

    G. Klambauer, T. Unterthiner, A. Mayr, and S. Hochreiter, “Self- normalizing neural networks,”Advances in neural information process- ing systems, vol. 30, 2017. AUTHORet al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS ON MEDICAL IMAGING 13

    2017

  26. [34]

    Vision transformer with deformable attention,

    Z. Xia, X. Pan, S. Song, L. E. Li, and G. Huang, “Vision transformer with deformable attention,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 4794–4803

    2022

  27. [35]

    The tumour microenviron- ment as an integrated framework to understand cancer biology,

    R. Burgos-Panadero, F. Lucantoni, E. Gamero-Sandemetrio, L. de la Cruz-Merino, T. ´Alvaro, and R. Noguera, “The tumour microenviron- ment as an integrated framework to understand cancer biology,”Cancer Letters, vol. 461, pp. 112–122, 2019

    2019

  28. [36]

    Study on density peaks clustering based on k-nearest neighbors and principal component analysis,

    M. Du, S. Ding, and H. Jia, “Study on density peaks clustering based on k-nearest neighbors and principal component analysis,”Knowledge- Based Systems, vol. 99, pp. 135–145, 2016

    2016

  29. [37]

    Dynam- icvit: Efficient vision transformers with dynamic token sparsification,

    Y . Rao, W. Zhao, B. Liu, J. Lu, J. Zhou, and C.-J. Hsieh, “Dynam- icvit: Efficient vision transformers with dynamic token sparsification,” Advances in neural information processing systems, vol. 34, pp. 13 937– 13 949, 2021

    2021

  30. [38]

    Distilling the knowledge in a neural network,

    G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,”arXiv preprint arXiv:1503.02531, 2015

    2015 arXiv

  31. [39]

    Transformer-based multimodal fusion for survival prediction by integrating whole slide images, clinical, and genomic data,

    Y . Chen, W. Zhao, and L. Yu, “Transformer-based multimodal fusion for survival prediction by integrating whole slide images, clinical, and genomic data,” in2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI). IEEE, 2023, pp. 1–5

    2023

  32. [40]

    Review the cancer genome atlas (tcga): an immeasurable source of knowledge,

    K. Tomczak, P. Czerwi ´nska, and M. Wiznerowicz, “Review the cancer genome atlas (tcga): an immeasurable source of knowledge,”Contem- porary Oncology/Wsp ´ołczesna Onkologia, vol. 2015, no. 1, pp. 68–77, 2015

    2015

  33. [41]

    An anatomic transcriptional atlas of human glioblastoma,

    R. B. Puchalskiet al., “An anatomic transcriptional atlas of human glioblastoma,”Science, vol. 360, no. 6389, pp. 660–663, 2018

    2018

  34. [42]

    Proteogenomic data and resources for pan-cancer analysis,

    Y . Liet al., “Proteogenomic data and resources for pan-cancer analysis,” Cancer cell, vol. 41, no. 8, pp. 1397–1406, 2023

    2023

  35. [43]

    Audio- visual fusion for sentiment classification using cross-modal autoen- coder,

    S. H. Dumpala, I. Sheikh, R. Chakraborty, and S. K. Kopparapu, “Audio- visual fusion for sentiment classification using cross-modal autoen- coder,” in32nd conference on neural information processing systems (NIPS 2018), 2019, pp. 1–4

    2018

  36. [44]

    Structure-preserving color normalization and sparse stain separation for histological images,

    A. Vahadaneet al., “Structure-preserving color normalization and sparse stain separation for histological images,”IEEE transactions on medical imaging, vol. 35, no. 8, pp. 1962–1971, 2016

    1962

  37. [45]

    Imagenet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchical image database,” in2009 IEEE conference on computer vision and pattern recognition. Ieee, 2009, pp. 248–255

    2009

  38. [46]

    Immport, toward repurposing of open access im- munological assay data for translational and clinical research,

    S. Bhattacharyaet al., “Immport, toward repurposing of open access im- munological assay data for translational and clinical research,”Scientific data, vol. 5, no. 1, pp. 1–9, 2018

    2018

  39. [47]

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

    A. Paszkeet al., “Pytorch: An imperative style, high-performance deep learning library,”Advances in neural information processing systems, vol. 32, 2019

    2019

  40. [48]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  41. [49]

    Morphological prototyping for unsupervised slide representation learning in computational pathology,

    A. H. Song, R. J. Chen, T. Ding, D. F. Williamson, G. Jaume, and F. Mahmood, “Morphological prototyping for unsupervised slide representation learning in computational pathology,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 11...

    2024

  42. [50]

    B. Mohanraj, “Glioblast: Establishing prognosis and targeted therapy for glioblastoma by applying convolutional neural networks to detect histological features, molecular subtypes, mgmt methylation, and egfr amplification from brain-biopsy whole-slide images.”

  43. [51]

    Decoding the interdependence of multiparametric magnetic resonance imaging to reveal patient subgroups correlated with survivals,

    C. Liet al., “Decoding the interdependence of multiparametric magnetic resonance imaging to reveal patient subgroups correlated with survivals,” Neoplasia, vol. 21, no. 5, pp. 442–449, 2019

    2019

Pith tools

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