Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Probabilistic Pretraining for Neural Regression

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The paper claims that probabilistic regression can be solved by one quantile-conditioned network pretrained on many tabular datasets, with positive transfer to new tasks and the largest gains under data scarcity.

desk verdict Transfer result is real, but the 'significantly outperforms trees' claim collapses against the paper's own appendix tables; worth reviewing, not desk-rejecting. read the letter →

arxiv 2508.16355 v1 pith:4QCCLNAA submitted 2025-08-22 cs.LG

Boris N. Oreshkin , Shiv Tavker , Dmitry Efimov This is my paper · ORCID
classification cs.LG
keywords NIAQUEprobabilisticregressionquantiletransferlearningtabulardatapermutationinvarianceCRPSfeatureimportance
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 introduces NIAQUE, a neural network that takes a tabular data row plus a requested quantile level and outputs that quantile of the target. Its central claim is that one such network can be pretrained jointly on many unrelated regression datasets and then fine-tuned on a new dataset, with positive transfer that is largest when labeled data is scarce. On a new 101-dataset benchmark, the globally pretrained model reports the best aggregate scores, ahead of local and global tree ensembles, a Transformer, TabPFN, and TabDPT. The paper also proves that the expected quantile loss is minimized only when the model outputs the inverse conditional cumulative distribution of the target, which is what makes the predicted quantiles meaningful. If these claims hold, tabular probabilistic regression stops being a per-dataset modeling chore and becomes a transfer-learning problem.

What carries the argument

The load-bearing mechanism is any-quantile learning: during training each example is paired with q drawn uniformly from (0,1) and scored with the pinball/quantile loss, and Theorem 1 ties this objective to CRPS and to the inverse CDF of the posterior. The architecture that makes transfer work is a prototype-aggregation encoder with dual residual loops: feature values and learned feature-ID embeddings are transformed per-feature, averaged into a fixed-size prototype, and residual blocks refine it while a delta constraint forces each block to encode only deviation from the current prototype. A FiLM-conditioned decoder turns the prototype plus any quantile vector into predictions. This separate

What would settle it

Run XGBoost-Local at depth 8, CatBoost-Local at depth 7, and LightGBM-Local with 50 leaves on the same 101-dataset protocol and compare SMAPE, RMSE, and CRPS to NIAQUE-Global; if the tuned trees match or beat it, the 'significantly outperforms' claim fails.

Watch

Extended reading notes

Core claim

NIAQUE's central claim is that probabilistic regression can be treated as a single multi-dataset learning problem. The model encodes each feature as a value plus a learned feature-ID embedding, aggregates the per-feature embeddings by averaging into a fixed-size prototype, and a decoder uses FiLM conditioning on the quantile q to emit quantile estimates. Because input dimension is handled by aggregation, datasets with different numbers of columns share one parameter set; the feature-ID embeddings act as an implicit task identifier. The paper proves Theorem 1: over a uniform q, the expected quantile loss equals a scaled CRPS and is minimized if and only if the predicted quantile function is t

Load-bearing premise

The claim that NIAQUE significantly outperforms tree-based methods rests on the tree hyperparameters used in the main comparison table; the paper's own appendix reports deeper XGBoost, CatBoost, and LightGBM models with aggregate scores close to NIAQUE's, so a fairer baseline choice could shrink the reported margin.

Editorial extensions

If this is right

  • A single pretrained model can be a reusable starting point for new regression tasks, so small-data practitioners no longer need to fit a model from scratch.
  • Full quantile curves support calibrated intervals, asymmetric error measures, and decision rules that need more than the median.
  • The linear O(d+Q) scaling per row makes the architecture practical for datasets with millions of rows, where attention-based tabular models hit memory limits.
  • Model-native feature importance is available at a fraction of the compute of post-hoc attribution, with high ranking agreement reported.
  • If transfer generalizes as reported, default practice for tabular regression could move from per-dataset tree tuning to pretrained neural regression plus light fine-tuning.

Reading between the lines

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

  • The feature-ID embeddings are described as an implicit task identifier; a direct test is to permute feature IDs for a held-out dataset before fine-tuning and see whether transfer degrades, separating learned task identity from feature semantics.
  • The marginal-confidence-interval importance construction only needs quantile outputs, so it could be applied to other probabilistic regressors, not just NIAQUE.
  • The benchmark samples all datasets uniformly; a curriculum or domain-weighted pretraining mix that emphasizes datasets similar to the target could amplify the reported transfer gains.
  • Because the model predicts arbitrary quantiles, inference-time quantile selection can adapt it to asymmetric costs or risk-averse decisions without retraining.
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

4 major / 5 minor

Summary. The paper introduces NIAQUE, a permutation-invariant encoder-decoder for probabilistic regression, and proposes to pretrain it jointly on many tabular regression datasets and fine-tune it on unseen tasks. Section 2.1 derives the any-quantile loss from the CRPS identity and proves (Theorem 1) that minimizing expected quantile loss recovers the inverse conditional CDF. The authors introduce TabRegSet-101, a 101-dataset regression benchmark, and report experiments on full-benchmark training, held-out transfer, and two Kaggle competitions. Claims include positive transfer from pretraining and significant superiority over tree-based models, Transformer baselines, TabPFN, and TabDPT.

Significance. If the claims held, the main contribution would be meaningful: scalable probabilistic transfer for tabular regression is underexplored, the architecture is more efficient than attention-based alternatives, and the paper ships a reusable benchmark and extensive ablations. The theoretical link between quantile loss and posterior CDFs is standard but cleanly presented, and the held-out transfer experiments (Table 2) show a consistent improvement direction. However, the headline empirical claim of significant superiority over tree-based models is not supported by the paper's own ablations. With the strongest settings the authors themselves report in Appendices F--H, XGBoost, CatBoost, and LightGBM all beat NIAQUE-Global on SMAPE and RMSE. The credible contribution after correcting the comparison is competitive point accuracy plus better probabilistic calibration and a transfer-learning gain, not state-of-the-art point prediction.

major comments (4)
  1. [Section 3.2, Table 1 vs Appendices F, G, H] The claim that NIAQUE-Global 'significantly outperforms both traditional tree-based methods' is contradicted by the paper's own hyperparameter sweeps when the strongest configurations are used. Table 10 reports XGBoost-Local at depth 8 with SMAPE 22.4, RMSE 0.773; Table 11 reports CatBoost-Local at depth 7 with SMAPE 21.5, RMSE 0.761, CRPS 0.272; Table 14 reports LightGBM-Local with 50 leaves at SMAPE 21.6, RMSE 0.752, CRPS 0.278. Against these, NIAQUE-Global (Table 1: SMAPE 22.1, RMSE 0.787, CRPS 0.261) is worse on SMAPE and RMSE and better only on CRPS and calibration. Table 1 instead reports depth-3 or leaf-limited baselines (XGBoost 25.6, CatBoost 24.3, LightGBM 25.7). The headline comparison must be rebuilt from the best validated configurations, or the claim should be narrowed to 'competitive on point metrics with better probabilistic calibration.'
  2. [Section 3.2, Table 2] The central positive-transfer result is presented without uncertainty quantification. Table 2 reports a single comparison per fine-tuning proportion ps, with no standard errors, confidence intervals, or paired significance tests. Appendix D.3 shows that 4-seed confidence intervals can be computed for the main benchmark, so the absence here is a choice. Since only 21 held-out datasets are used, the consistent but unquantified gap between NIAQUE-Pretrain and NIAQUE-Scratch could depend on a small number of datasets. Please add bootstrap or paired intervals, or otherwise quantify the variability.
  3. [Section 2.1 and Appendix B, Theorem 1] The theorem statement is mathematically imprecise. It defines 'F as a probability measure over variable y' and concludes the minimizer of E ρ(y,F^{-1}(q)) is F = P_{y|x}. If F is a single unconditional distribution independent of x, the minimizer is the marginal P_y, not P_{y|x}; the proof only goes through if F is allowed to depend on x, i.e., F = (F_x)_{x∈X} with F_x^{-1}(q) corresponding to fθ(x,q). Please restate the theorem with conditional distributions F_x and clarify that the minimization is pointwise in x. The underlying standard result is correct, but the current statement is not.
  4. [Section 3.2 and Appendix D.2] The Kaggle tree-based baseline rows are adopted from public notebooks and forum posts rather than being run under the authors' own controlled protocol (see Table 7 caption: 'Baseline results are adopted from publicly shared notebooks and discussion forums'). These external results may use different hardware, preprocessing, augmentation, or ensembling. The statement that NIAQUE 'outperforms vanilla tree-based models' in Kaggle settings is therefore not a controlled comparison. The authors should either run the tree baselines under the same protocol or explicitly qualify the claim as 'competitive with published public solutions.'
minor comments (5)
  1. [Table 3 and Appendix D.2.1] Table 3 labels the second NIAQUE row simply 'NIAQUE-Pretrain', while Appendix D.2.1 distinguishes NIAQUE-Pretrain-100 (Abalone excluded) from NIAQUE-Pretrain-full (Abalone included). Add a footnote in Table 3 to make clear which variant is reported.
  2. [Abstract and Section 1] The phrase 'closes this gap' and 'for the first time' are stronger than the evidence warrants; transfer learning for regression has been studied in related forms. Suggest softening.
  3. [Section 2.4] The 'implicit task ID' hypothesis is presented as the mechanism for transfer, but the supporting evidence is only the qualitative UMAP visualization in Fig. 2. A quantitative analysis (e.g., embedding similarity vs task similarity) would strengthen the claim.
  4. [Section 3.1 and Appendix D.3] Confidence intervals are reported only for a subset of models in Table 9; the main Table 1 lacks them for the full set. Please include CIs for all rows or state why some are omitted.
  5. [Appendix H, Table 14] The LightGBM max_depth=2 rows show identical metrics across all leaf counts, which suggests the depth cap dominates and the leaf parameter is inactive. This should be noted when interpreting those baseline settings.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the theoretical result is derived from external CRPS identities and transfer is evaluated on genuinely held-out data, with routine self-citations only.

full rationale

No circular derivation is present in the paper's load-bearing chain. Theorem 1 is proved by combining the standard CRPS identity CRPS(F,y)=2∫ρ(y,F^{-1}(q))dq (Eq. 9, cited to Gneiting and Ranjan 2011) with the L2 representation of CRPS (Eq. 2) and completing the square; the minimizer is computed, not assumed. The subsequent claims about SGD convergence and the inversion-method interpretation cite Farrell et al. (2021) and Devroye (1986), both external to the authors. Transfer learning is evaluated on 21 datasets held out from the 80 pretraining datasets (Table 2), and the Kaggle Abalone study explicitly reports NIAQUE-Pretrain-100, which excludes the original UCI Abalone dataset from pretraining, so the transfer gain is not forced by construction. The self-citations to Smyl et al. (2024) and Oreshkin et al. (2022) describe architectural lineage and are not used as evidence for the paper's theoretical or empirical claims; the architecture equations are stated in full in Section 2.2. The separate concern that Table 1 uses weaker tree hyperparameters than the paper's own appendix sweeps (e.g., CatBoost-Local depth 7 in Table 11 achieves SMAPE 21.5 and RMSE 0.761 versus NIAQUE-Global's 22.1 and 0.787) is a benchmark-fairness and correctness issue, not circularity: no fitted parameter is renamed as a prediction, and no claim reduces by definition to its inputs. The score of 1 reflects only the presence of routine self-citations that are not load-bearing.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

NIAQUE's headline numbers depend on hand-chosen architecture hyperparameters (E=1024, Ein=64, R=4, L=2, dp=0.2), the 5% single-feature augmentation, per-dataset target rescaling to [0,10], and, critically, on which hyperparameters are reported for the tree baselines in Table 1 versus the stronger settings in the appendix sweeps. The theory rests on four imported results, the least justified being the 'implicit task ID' transfer hypothesis (Section 2.4) and the assumption that sample-level aggregation across 101 rescaled datasets is a meaningful metric (Appendix D shows the same aggregation breaks the confidence-interval computation). No new physical or latent entities are introduced.

free parameters (4)
  • Architecture and training hyperparameters (E, Ein, R, L, dp, lr, batch) = E=1024, Ein=64, R=4, L=2, dp=0.2, lr=1e-4, batch=512
    Selected on validation within the authors' own TabRegSet-101; the ablation (Appendix J) shows sensitivity (2 blocks, width 64: SMAPE 24.2). These choices are not derived from the theory.
  • Single-feature rows fraction = 5% of training rows
    Added to enable CI-based feature importance (Section 2.3); Appendix K shows interpretability fails without them. Chosen by hand.
  • Per-dataset target normalization to [0,10] = target rescaled per dataset to [0,10]
    Chosen to equalize metrics across datasets (Appendix C); every aggregate number in Tables 1-2 depends on this scale, and the feature log-transform Eq. (10) similarly conditions all inputs.
  • Table 1 tree baseline settings = XGBoost max_depth 3, CatBoost depth 3, LightGBM max_depth 2, num_leaves 5
    These are the weakest settings of the paper's own sweeps (Appendix F/G/H); stronger settings match or beat NIAQUE on point accuracy. The choice is load-bearing for the superiority claim.
assumptions (5)
  • standard math Farrell et al. (2021) Theorem 1: SGD-trained MLPs with width/depth growing with sample size converge to the risk minimizer
    Invoked in Section 2.1 to conclude the trained NIAQUE outputs the inverse conditional CDF; NIAQUE is a fixed 4-block 1024-wide network, so the scaling conditions are unverified.
  • standard math CRPS-pinball identity (Gneiting and Ranjan 2011), Eq. (9)
    Foundation of Theorem 1; classical and correct.
  • standard math Inversion method (Devroye 1986, Thm 2.1)
    Used in Section 2.1 to interpret uniform-q predictions as posterior samples.
  • ad hoc to paper 'Implicit task ID' hypothesis: shared feature-ID embeddings and prototype pooling transfer across unrelated datasets
    Section 2.4 states this as the transfer mechanism; it is the core modeling premise of the transfer claim, presented without isolation or negative-transfer analysis.
  • domain assumption Aggregate sample-level evaluation across rescaled datasets is a meaningful benchmark metric
    Table 1 aggregates test points across 101 datasets after [0,10] rescaling, weighting large datasets; Appendix D.1 shows the same aggregation produces invalid confidence intervals, questioning its statistical validity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Probabilistic Pretraining for Neural Regression." pith.science (2026). https://pith.science/paper/4QCCLNAA

@misc{pith2026250816355,
  author       = {Pith},
  title        = {Pith review of: Probabilistic Pretraining for Neural Regression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4QCCLNAA}},
  note         = {Machine review of arXiv:2508.16355}
}
read the original abstract

Transfer learning for probabilistic regression remains underexplored. This work closes this gap by introducing NIAQUE, Neural Interpretable Any-Quantile Estimation, a new model designed for transfer learning in probabilistic regression through permutation invariance. We demonstrate that pre-training NIAQUE directly on diverse downstream regression datasets and fine-tuning it on a specific target dataset enhances performance on individual regression tasks, showcasing the positive impact of probabilistic transfer learning. Furthermore, we highlight the effectiveness of NIAQUE in Kaggle competitions against strong baselines involving tree-based models and recent neural foundation models TabPFN and TabDPT. The findings highlight NIAQUE's efficacy as a robust and scalable framework for probabilistic regression, leveraging transfer learning to enhance predictive performance.

Figures

Figures reproduced from arXiv: 2508.16355 by the authors.

Figure 1
Figure 1. NIAQUE’s encoder-decoder architecture transforms variable-dimensional inputs into fixed [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. UMAP projections of embeddings derived from NIAQUE’s feature encoder for each sample, [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Statistics of the evaluation dataset: (a) distribution by source, (b) dataset sizes, and (c) [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Transformer baseline used in our experiments. The feature encoding module is replaced [PITH_FULL_IMAGE:figures/full_fig_p031_4.png]
Figure 5
Figure 5. Figure 5: Training losses with (dark red) and without (blue) input value log-transform eq. (10). The [PITH_FULL_IMAGE:figures/full_fig_p035_5.png]
Figure 6
Figure 6. Figure 6: The effect of adding training rows containing only one of the input features as NIAQUE [PITH_FULL_IMAGE:figures/full_fig_p035_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. AGNFormer I: Reconstruction of AGN spectra using a probabilistic transformer model

    astro-ph.GA 2026-07 conditional novelty 6.0 of 10

    An uncertainty-aware transformer reconstructs masked AGN broad lines and spectral halves with 4-16% flux errors and beats eleven purpose-built Lyα-reconstruction algorithms on a blind benchmark.

Reference graph

Works this paper leans on

58 extracted references · 50 canonical work pages · cited by 1 Pith paper

  1. [1]

    KEEL data-mining software tool: Data set repository, integration of algorithms and experimental analysis framework

    Jesús Alcalá-Fdez, Alberto Fernández, Julián Luengo, Joaquín Derrac, and Salvador García. KEEL data-mining software tool: Data set repository, integration of algorithms and experimental analysis framework. J. Multiple Valued Log. Soft Comput., 17 0 (2--3): 0 255--287, 2011

    work page 2011

  2. [2]

    1st place solution for the regression with a flood prediction dataset

    Aldparis. 1st place solution for the regression with a flood prediction dataset. https://www.kaggle.com/competitions/playground-series-s4e5/discussion/509043, 2024. Kaggle

    work page 2024

  3. [3]

    Maddix, Michael W

    Abdul Fatir Ansari, Lorenzo Stella, Caner Turkmen, Xiyuan Zhang, Pedro Mercado, Huibin Shen, Oleksandr Shchur, Syama Syndar Rangapuram, Sebastian Pineda Arango, Shubham Kapoor, Jasper Zschiegner, Danielle C. Maddix, Michael W. Mahoney, Kari Torkkola, Andrew Gordon Wilson, Michael Bohlke-Schneider, and Yuyang Wang. Chronos: Learning the language of time se...

    arXiv 2024

  4. [4]

    Arik and Tomas Pfister

    Sercan Ö. Arik and Tomas Pfister. Tabnet: Attentive interpretable tabular learning. In Proc. AAAI, pages 6679--6687, May 2021

    work page 2021

  5. [5]

    House rent prediction dataset, 2022

    Sourav Banerjee. House rent prediction dataset, 2022. URL https://www.kaggle.com/datasets/iamsouravbanerjee/house-rent-prediction-dataset

    work page 2022

  6. [6]

    Random forests

    Leo Breiman. Random forests. Machine Learning, 45 0 (1): 0 5--32, 2001

    2001

  7. [7]

    Xgboost: A scalable tree boosting system

    Tianqi Chen and Carlos Guestrin. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’16. ACM, August 2016. doi:10.1145/2939672.2939785. URL http://dx.doi.org/10.1145/2939672.2939785

    arXiv 2016

  8. [8]

    ps-s4e4: Eda | lightgbm | importance plots

    Gourab Roy (dataWr3cker). ps-s4e4: Eda | lightgbm | importance plots. https://www.kaggle.com/code/gourabr0y555/ps-s4e4-eda-lightgbm-importance-plots, 2024. Kaggle

    work page 2024

Show all 58 references
  1. [9]

    Ames, iowa: Alternative to the boston housing data as an end of semester regression project

    Dean De Cock. Ames, iowa: Alternative to the boston housing data as an end of semester regression project. Journal of Statistics Education, 19 0 (3), 2011

    2011

  2. [10]

    BERT : Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT : Pre-training of deep bidirectional transformers for language understanding. In Jill Burstein, Christy Doran, and Thamar Solorio, editors, NAACL-HLT (1), pages 4171--4186. Association for Computational Lin...

    2019

  3. [11]

    Non-Uniform Random Variate Generation

    Luc Devroye. Non-Uniform Random Variate Generation. Springer-Verlag, New York, NY, USA, 1986

    1986

  4. [12]

    Flood prediction factors

    Brijlal Dhankour. Flood prediction factors. https://www.kaggle.com/datasets/brijlaldhankour/flood-prediction-factors, 2024. Kaggle

    2024

  5. [13]

    Uci machine learning repository: Abalone data set, 2019

    Dheeru Dua and Casey Graff. Uci machine learning repository: Abalone data set, 2019. URL https://archive.ics.uci.edu/ml/datasets/abalone

    2019

  6. [14]

    Deep neural networks for estimation and inference

    Max H Farrell, Tengyuan Liang, and Sanjog Misra. Deep neural networks for estimation and inference. Econometrica, 89 0 (1): 0 181--213, January 2021

    2021

  7. [15]

    Marta Garnelo, Dan Rosenbaum, Christopher Maddison, Tiago Ramalho, David Saxton, Murray Shanahan, Yee Whye Teh, Danilo Rezende, and S. M. Ali Eslami. Conditional neural processes. In Proc. ICML, volume 80, pages 1704--1713. PMLR, Jul 2018 a . URL https://proceedings.mlr.press/...

    2018

  8. [16]

    Rezende, S

    Marta Garnelo, Jonathan Schwarz, Dan Rosenbaum, Fabio Viola, Danilo J. Rezende, S. M. Ali Eslami, and Yee Whye Teh. Neural processes, 2018 b . URL https://arxiv.org/abs/1807.01622

    2018 arXiv

  9. [17]

    Timegpt-1, 2023

    Azul Garza and Max Mergenthaler-Canseco. Timegpt-1, 2023

    2023

  10. [18]

    Comparing density forecasts using threshold-and quantile-weighted scoring rules

    Tilmann Gneiting and Roopesh Ranjan. Comparing density forecasts using threshold-and quantile-weighted scoring rules. Journal of Business & Economic Statistics, 29 0 (3): 0 411--422, 2011

    2011

  11. [19]

    Why do tree-based models still outperform deep learning on typical tabular data? In Proc

    Leo Grinsztajn, Edouard Oyallon, and Gael Varoquaux. Why do tree-based models still outperform deep learning on typical tabular data? In Proc. NeurIPS, pages 507--520, 2022

    2022

  12. [20]

    1st place solution for the regression with an abalone dataset competition

    Johannes Heller. 1st place solution for the regression with an abalone dataset competition. https://www.kaggle.com/competitions/playground-series-s4e4/discussion/499174, 2024. Kaggle

    2024

  13. [21]

    Tab PFN : A transformer that solves small tabular classification problems in a second

    Noah Hollmann, Samuel M \"u ller, Katharina Eggensperger, and Frank Hutter. Tab PFN : A transformer that solves small tabular classification problems in a second. In Proc. ICLR, 2023

    2023

  14. [22]

    TabTransformer : Tabular data modeling using contextual embeddings

    Xin Huang, Ashish Khetan, Milan Cvitkovic, and Zohar Karnin. TabTransformer : Tabular data modeling using contextual embeddings. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 7671--7679. AAAI Press, 2021

    2021

  15. [23]

    Well-tuned simple nets excel on tabular datasets

    Arlind Kadra, Marius Lindauer, Frank Hutter, and Josif Grabocka. Well-tuned simple nets excel on tabular datasets. In Proc. NeurIPS, 2021

    2021

  16. [24]

    Kaggle datasets, 2024

    Kaggle. Kaggle datasets, 2024. URL https://www.kaggle.com/datasets

    2024

  17. [25]

    Linear convergence of gradient and proximal-gradient methods under the polyak-lojasiewicz condition

    Hamed Karimi, Julie Nutini, and Mark Schmidt. Linear convergence of gradient and proximal-gradient methods under the polyak-lojasiewicz condition. In Paolo Frasconi, Niels Landwehr, Giuseppe Manco, and Jilles Vreeken, editors, Machine Learning and Knowledge Discovery in Databa...

    2016

  18. [26]

    Lightgbm: A highly efficient gradient boosting decision tree

    Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. Lightgbm: A highly efficient gradient boosting decision tree. In I. Guyon, U. Von Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors, Advances in ...

    2017

  19. [27]

    UCI machine learning repository, 2017

    Markelle Kelly, Rachel Longjohn, and Kolby Nottingham. UCI machine learning repository, 2017. URL http://archive.ics.uci.edu/ml

    2017

  20. [28]

    Transfer learning with deep tabular models

    Roman Levin, Valeriia Cherepanova, Avi Schwarzschild, Arpit Bansal, C Bayan Bruss, Tom Goldstein, Andrew Gordon Wilson, and Micah Goldblum. Transfer learning with deep tabular models. In Proc. ICLR, 2023

    2023

  21. [29]

    Tabdpt: Scaling tabular foundation models

    Junwei Ma, Valentin Thomas, Rasa Hosseinzadeh, Hamidreza Kamkari, Alex Labach, Jesse C Cresswell, Keyvan Golestan, Guangwei Yu, Maksims Volkovs, and Anthony L Caterini. Tabdpt: Scaling tabular foundation models. arXiv preprint arXiv:2410.18164, 2024

    2024

  22. [30]

    Xgb|cat|lightgbm

    Ali Masoudi. Xgb|cat|lightgbm. https://www.kaggle.com/code/alimsd77/xgb-cat-lightgbm#lightgbm, 2024. Kaggle

    2024

  23. [31]

    Amazon price and product data: Electronic commerce, 2015

    Julian McAuley, Rahul Pandey, and Jure Leskovec. Amazon price and product data: Electronic commerce, 2015. URL https://www.kaggle.com/datasets/skillsmuggler/amazon-products-dataset

    2015

  24. [32]

    McInnes , J

    L. McInnes , J. Healy , and J. Melville . UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction . ArXiv e-prints, February 2018

    2018

  25. [33]

    Baseline v1 | catboost

    Jay Milind. Baseline v1 | catboost. https://www.kaggle.com/code/jaymilindpadloskar/baseline-v1-catboost, 2024. Kaggle

    2024

  26. [34]

    Transformers can do Bayesian inference

    Samuel M \"u ller, Noah Hollmann, Sebastian Pineda Arango, Josif Grabocka, and Frank Hutter. Transformers can do Bayesian inference. In Proc. ICLR, 2022

    2022

  27. [35]

    TabPFN: A Transformer that solves small tabular classification problems in a second

    Felix Müller, Christian Holtz, Emir Akyürek, Lukas Lorbeer, and Frank Hutter. TabPFN: A Transformer that solves small tabular classification problems in a second . https://github.com/PriorLabs/TabPFN, 2025. URL https://github.com/PriorLabs/TabPFN

    2025

  28. [36]

    Occupancy flow: 4d reconstruction by learning particle dynamics

    Michael Niemeyer, Lars Mescheder, Michael Oechsle, and Andreas Geiger. Occupancy flow: 4d reconstruction by learning particle dynamics. In Proc. ICCV, October 2019

    2019

  29. [37]

    Olson, William La Cava, Patryk Orzechowski, Ryan J

    Randal S. Olson, William La Cava, Patryk Orzechowski, Ryan J. Urbanowicz, and Jason H. Moore. Pmlb: A large benchmark suite for machine learning evaluation and comparison. BioData Mining, 10 0 (36), 2017 a

    2017

  30. [38]

    Olson, William La Cava, Patryk Orzechowski, Ryan J

    Randal S. Olson, William La Cava, Patryk Orzechowski, Ryan J. Urbanowicz, and Jason H. Moore. Pmlb: a large benchmark suite for machine learning evaluation and comparison. BioData Mining, 10 0 (1): 0 36, Dec 2017 b . ISSN 1756-0381. doi:10.1186/s13040-017-0154-4. URL https://d...

    2017 doi

  31. [39]

    Wind dataset, n.d

    OpenML. Wind dataset, n.d. URL https://www.openml.org/data/download/52615/wind.arff

  32. [40]

    Oreshkin, Florent Bocquelet, F \' e lix G

    Boris N. Oreshkin, Florent Bocquelet, F \' e lix G. Harvey, Bay Raitt, and Dominic Laflamme. Protores: Proto-residual network for pose authoring via learned inverse kinematics. In Proc. ICLR, 2022

    2022

  33. [41]

    Film: Visual reasoning with a general conditioning layer

    Ethan Perez, Florian Strub, Harm De Vries, Vincent Dumoulin, and Aaron Courville. Film: Visual reasoning with a general conditioning layer. In Proc. AAAI, 2018

    2018

  34. [42]

    Catboost: unbiased boosting with categorical features, 2019

    Liudmila Prokhorenkova, Gleb Gusev, Aleksandr Vorobev, Anna Veronika Dorogush, and Andrey Gulin. Catboost: unbiased boosting with categorical features, 2019

    2019

  35. [43]

    Qi, Hao Su, Kaichun Mo, and L

    C. Qi, Hao Su, Kaichun Mo, and L. Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. Proc. CVPR, pages 77--85, 2017

    2017

  36. [44]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. In Proc. ICML, volum...

    2021

  37. [45]

    Dai, Nissan Hajaj, Michaela Hardt, Peter J

    Alvin Rajkomar, Eyal Oren, Kai Chen, Andrew M. Dai, Nissan Hajaj, Michaela Hardt, Peter J. Liu, Xiaobing Liu, Jake Marcus, Marvin Sun, et al. Scalable and accurate deep learning with electronic health records. Nature Medicine, 24 0 (7): 0 1337--1340, 2018

    2018

  38. [46]

    Regression with an abalone dataset

    Walter Reade and Ashley Chow. Regression with an abalone dataset. https://kaggle.com/competitions/playground-series-s4e4, 2024 a . Kaggle

    2024

  39. [47]

    Regression with a flood prediction dataset

    Walter Reade and Ashley Chow. Regression with a flood prediction dataset. https://www.kaggle.com/competitions/playground-series-s4e5, 2024 b . Kaggle

    2024

  40. [48]

    Pmlb v1.0: an open source dataset collection for benchmarking machine learning methods

    Joseph D Romano, Trang T Le, William La Cava, John T Gregg, Daniel J Goldberg, Praneel Chakraborty, Natasha L Ray, Daniel Himmelstein, Weixuan Fu, and Jason H Moore. Pmlb v1.0: an open source dataset collection for benchmarking machine learning methods. arXiv preprint arXiv:20...

    2012 arXiv

  41. [49]

    Flood forecasting with xgboost

    Zeyad Sayed. Flood forecasting with xgboost. https://www.kaggle.com/code/zeyadsayedadbullah/flood-forecasting-with-xgboost, 2024. Kaggle

    2024

  42. [50]

    Why nns is better than gbds? https://www.kaggle.com/competitions/playground-series-s4e4/discussion/496471\#2767909, 2024

    Broccoli Beef (siukeitin). Why nns is better than gbds? https://www.kaggle.com/competitions/playground-series-s4e4/discussion/496471\#2767909, 2024. Kaggle

    2024

  43. [51]

    Oreshkin, Paweł Pełka, and Grzegorz Dudek

    Slawek Smyl, Boris N. Oreshkin, Paweł Pełka, and Grzegorz Dudek. Any-quantile probabilistic forecasting of short-term electricity demand, 2024. URL https://arxiv.org/abs/2404.17451

    2024 arXiv

  44. [52]

    Jake Snell, Kevin Swersky, and Richard S. Zemel. Prototypical networks for few-shot learning. In Proc. NIPS, pages 4080--4090, 2017

    2017

  45. [53]

    Monocular, One-stage, Regression of Multiple 3D People

    Yu Sun, Qian Bao, Wu Liu, Yili Fu, Black Michael J., and Tao Mei. Monocular, One-stage, Regression of Multiple 3D People . In ICCV, 2021

    2021

  46. [54]

    van Rijn, Bernd Bischl, and Luis Torgo

    Joaquin Vanschoren, Jan N. van Rijn, Bernd Bischl, and Luis Torgo. OpenML : networked science in machine learning. SIGKDD Explorations, 15 0 (2): 0 49--60, 2013. doi:10.1145/2641190.2641198. URL http://doi.acm.org/10.1145/2641190.264119

    2013

  47. [55]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. In I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors, Proc. NeurIPS, volume...

    2017

  48. [56]

    S4e4 | abalone | catboost

    Suraj Wate. S4e4 | abalone | catboost. https://www.kaggle.com/code/surajwate/s4e4-abalone-catboost, 2024. Kaggle

    2024

  49. [57]

    Deep sets

    Manzil Zaheer, Satwik Kottur, Siamak Ravanbakhsh, Barnabas Poczos, Russ R Salakhutdinov, and Alexander J Smola. Deep sets. In Proc. NeurIPS, volume 30. Curran Associates, Inc., 2017

    2017

  50. [58]

    Openfe: automated feature generation with expert-level performance

    Tianping Zhang, Zheyu Zhang, Zhiyuan Fan, Haoyan Luo, Fengyuan Liu, Qian Liu, Wei Cao, and Jian Li. Openfe: automated feature generation with expert-level performance. In Proc. ICML, 2023

    2023

Pith tools

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