pith. sign in

arxiv: 2606.25188 · v1 · pith:ZVQS5J47new · submitted 2026-06-23 · 💻 cs.LG

Efficient Analytic Uncertainty Quantification for Multi-Modal Regression

Kun Jin , James Harrison , Jiawei Li , Sihan Liu , Jiayi Liu , Randolph Linderman , Yuening Li , Arnab Bhadury
show 3 more authors
Sourabh Prakash Bansod Liang Liu Jasper Snoek
This is my paper

Pith reviewed 2026-06-25 23:37 UTC · model grok-4.3

classification 💻 cs.LG
keywords uncertainty quantificationmulti-modal regressionvariational Bayesian inferencequantile regressionclassification restorationepistemic uncertaintyactive learning
0
0 comments X

The pith

Quantile regression and classification restoration admit variational Bayesian formulations that produce analytic evidence lower bounds and predictive densities for efficient uncertainty quantification in multi-modal regression tasks.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper seeks to deliver efficient uncertainty quantification for regression problems whose label distributions exhibit multiple modes. It embeds quantile regression and classification restoration inside variational Bayesian inference so that both training and inference admit analytic or easily approximated expressions. This removes the need to assume single-peak forms such as Gaussians while avoiding the computational cost of ensembles. A sympathetic reader would care because many real regression tasks produce complex conditional densities yet still require trustworthy variance estimates at scale. If the formulations hold, accurate multi-modal modeling and fast uncertainty estimates become simultaneously available.

Core claim

The central claim is that novel formulations of quantile regression and classification restoration inside the variational Bayesian inference framework yield analytic evidence lower bounds for training and closed-form or analytically approximated predictive densities for inference, thereby achieving accurate estimation of complex conditional distributions together with highly efficient uncertainty quantification.

What carries the argument

The unified distribution-agnostic variational Bayesian inference framework that reformulates quantile regression and classification restoration to admit analytic ELBOs and predictive densities.

If this is right

  • The method outperforms state-of-the-art multi-modal regression baselines on three large-scale benchmarks with multi-modal label distributions.
  • Predictive performance matches that of computationally expensive ensemble models.
  • Epistemic uncertainty estimates enable highly data-efficient active learning strategies.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same reformulation strategy could be applied to other semi-parametric density estimators that currently lack analytic uncertainty.
  • The closed-form predictive densities may reduce the cost of propagating uncertainty through downstream decision pipelines.
  • Because the approach is distribution-agnostic, it offers a route to uncertainty quantification in settings where label supports change over time.

Load-bearing premise

Quantile regression and classification restoration can be placed inside variational Bayesian inference so that the resulting evidence lower bounds and predictive densities remain analytic or cheaply approximated without substantial loss of multi-modal fidelity.

What would settle it

On a synthetic multi-modal regression dataset, compare the analytic predictive density against exact Monte Carlo sampling from the true conditional and check whether the total variation distance or negative log-likelihood gap exceeds a threshold that materially changes downstream active-learning performance.

Figures

Figures reproduced from arXiv: 2606.25188 by Arnab Bhadury, James Harrison, Jasper Snoek, Jiawei Li, Jiayi Liu, Kun Jin, Liang Liu, Randolph Linderman, Sihan Liu, Sourabh Prakash Bansod, Yuening Li.

Figure 1
Figure 1. Figure 1: Visualize our contributions with synthetic data. The target distribution is bimodal (y has high probability mass around 0 and 1), which triggers the “Ghost Value” pathology at y ≈ 0.5 in unimodal models (Panel A). The x > 2 region denotes out-of-distribution (OOD) inputs where the ground truth distribution is gradually transitioning into pure noise. Our framework achieves (1) representation of the conditio… view at source ↗
Figure 2
Figure 2. Figure 2: The Modular Distribution-Agnostic Uncertainty Framework. Both paths share a distance-aware backbone and variational inference engine, but diverge at the output head to address different topological needs: (a) QR-VBLL is better at continuous, smooth shapes, since it excels at Global Fidelity, smoothly capturing the gradual spread without binning artifacts and (b) CR-VBLL is better at sharp, disjoint densiti… view at source ↗
Figure 3
Figure 3. Figure 3: Target distributions for Real-World Benchmarks. The label density plots reveal the complex, [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: The test NLL curves for active learning methods on KuaiRec, demon￾strating data efficiency improvement. Moreover, both CR-VBLL and QR-VBLL are having a much larger advantage margin from the “safety net” effect (like [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Analytic smoothing of the Asymmetric Laplace surrogate likelihood. The standard [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: CR-VBLL Sensitivity (Bins KCR): Efficiency vs. Calibration Trade-off. While increasing KCR from 20 to 40 yields significant NLL gains across all datasets (Left), further increasing KCR often leads to CRPS degradation (Right) due to the “sharpening effect,” where the model becomes overconfident in specific bins. We select KCR = 40 as the optimal balance between predictive likelihood and calibration stabilit… view at source ↗
Figure 8
Figure 8. Figure 8: QR-VBLL Sensitivity (Quantiles KQR): High-Resolution Robustness. Unlike CR, Quantile Regression benefits monotonically from higher resolution. On the sparse Uber dataset, NLL improves continuously as KQR increases (Left), while calibration (CRPS) remains stable without degradation (Right). We select KQR = 100 to maximize precision on difficult distributions without incurring penalties on simpler ones. Here… view at source ↗
Figure 9
Figure 9. Figure 9: Reliability Diagram on WeChat. Comparison of calibration curves between Vanilla Gaussian regression and our CR-VBLL method. The Gaussian model shows significant deviation due to the unimodal assumption failure, while CR-VBLL maintains robust calibration (hugging the diagonal), effectively acting as a “Safety Net” by correctly modeling the multi-modal density. 31 [PITH_FULL_IMAGE:figures/full_fig_p031_9.png] view at source ↗
read the original abstract

Efficient uncertainty quantification (UQ) is essential for trustworthy large-scale learning. Existing UQ methods for regression tasks mainly operate under the assumption that the conditional label marginal satisfies single-peak parametric models, e.g., Gaussians, where the negative log-likelihood function simplifies to the mean square error. However, such single-peak assumptions fail in regression tasks featuring multi-modal distributions. On the other hand, semi-parametric methods which achieve strong regression performance for multi-modal distributions often lack efficient quantification on their prediction variances. In this work, we extend UQ techniques based on Variational Bayesian Inference (VBI) to two widely used semi-parametric regression models that yield histogram-like reconstructions of the conditional label densities: Quantile Regression (QR) and Classification Restoration (CR). Our approach introduces a unified, distribution-agnostic framework that simultaneously achieves accurate estimation of complex conditional distributions and highly efficient UQ. Theoretically, our method is grounded in novel formulations of QR and CR within the VBI framework, yielding analytic Evidence Lower Bounds (ELBO) to streamline training and a closed-form or analytically approximated predictive density for efficient inference. Empirically, we evaluate our methods on three large-scale regression benchmarks with multi-modal label distributions. Our framework outperforms state-of-the-art multi-modal regression baselines, and even matches predictive performance of computationally expensive ensemble models. Furthermore, by leveraging epistemic uncertainty estimation, our approach enables highly data-efficient active learning strategies.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The paper extends variational Bayesian inference to quantile regression and classification restoration to handle multi-modal regression, claiming novel formulations that produce analytic ELBOs for training and closed-form or analytically approximated predictive densities for efficient UQ. It reports empirical outperformance over multi-modal baselines on three large-scale benchmarks and parity with ensembles, plus benefits for active learning via epistemic uncertainty.

Significance. If the analytic ELBO and predictive-density claims hold without hidden approximations that compromise multi-modality or distribution-agnosticism, the framework would supply a computationally lightweight UQ method for complex conditional distributions, with direct applicability to data-efficient active learning.

major comments (1)
  1. [Abstract] Abstract: the assertion of 'analytic Evidence Lower Bounds (ELBO)' and 'closed-form or analytically approximated predictive density' is presented without any explicit ELBO expression, derivation outline, variational family, or reparameterization. Because these properties are load-bearing for the central claim of efficient, distribution-agnostic UQ, their absence makes it impossible to assess whether the formulations are truly analytic or introduce fidelity loss.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful reading and constructive feedback on our manuscript. We address the single major comment below and will revise the abstract accordingly to improve clarity while preserving the paper's core claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the assertion of 'analytic Evidence Lower Bounds (ELBO)' and 'closed-form or analytically approximated predictive density' is presented without any explicit ELBO expression, derivation outline, variational family, or reparameterization. Because these properties are load-bearing for the central claim of efficient, distribution-agnostic UQ, their absence makes it impossible to assess whether the formulations are truly analytic or introduce fidelity loss.

    Authors: We agree that the abstract would benefit from greater specificity to allow readers to immediately assess the analytic claims. The explicit ELBO derivations, variational family (a product of independent quantile or class-conditional distributions under a mean-field assumption), and reparameterization details appear in Sections 3.2–3.3 (QR-VBI) and 4.2–4.3 (CR-VBI). These yield closed-form ELBO terms for the chosen families without additional approximations that would compromise multi-modality. In the revised manuscript we will expand the abstract to briefly name the variational family and direct readers to the relevant sections, thereby addressing the concern without altering the distribution-agnostic character of the framework. revision: yes

Circularity Check

0 steps flagged

No circularity: novel VBI formulations of QR/CR presented as independent extensions yielding analytic ELBOs

full rationale

The paper's central claim rests on introducing novel formulations of quantile regression and classification restoration inside the variational Bayesian inference framework that produce analytic ELBOs and closed-form or approximated predictive densities. No equations, derivations, or self-citations are supplied in the provided text that reduce any claimed prediction or ELBO to a fitted parameter or prior result by construction. The description treats the VBI extension as supplying independent analytic structure rather than renaming or refitting existing quantities. This matches the default expectation of a self-contained derivation against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that QR and CR produce histogram-like conditional densities compatible with analytic VBI treatment; no free parameters or invented entities are identifiable from the abstract alone.

axioms (1)
  • domain assumption Quantile regression and classification restoration yield histogram-like reconstructions of conditional label densities that admit analytic or approximable ELBOs under VBI.
    This premise is required for the unified framework and analytic results described in the abstract.

pith-pipeline@v0.9.1-grok · 5811 in / 1127 out tokens · 29594 ms · 2026-06-25T23:37:52.284240+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

76 extracted references · 3 canonical work pages · 1 internal anchor

  1. [1]

    Bayesian Analysis , year=

    Bayesian regularized quantile regression , author=. Bayesian Analysis , year=

  2. [2]

    Journal of statistical computation and simulation , _volume=

    Bayesian quantile inference , author=. Journal of statistical computation and simulation , _volume=. 2003 , _publisher=

    2003

  3. [3]

    Journal of Money, Credit and Banking , _volume=

    Constructing density forecasts from quantile regressions , author=. Journal of Money, Credit and Banking , _volume=. 2012 , _publisher=

    2012

  4. [4]

    2005 , publisher=

    Quantile regression , author=. 2005 , publisher=

    2005

  5. [5]

    International Statistical Review , _volume=

    Posterior inference in Bayesian quantile regression with asymmetric Laplace likelihood , author=. International Statistical Review , _volume=. 2016 , _publisher=

    2016

  6. [6]

    Journal of statistical computation and simulation , _volume=

    Gibbs sampling methods for Bayesian quantile regression , author=. Journal of statistical computation and simulation , _volume=. 2011 , _publisher=

    2011

  7. [7]

    Under review , year=

    Heteroscedastic Variational Last Layers , author=. Under review , year=

  8. [8]

    Econometrica , _pages=

    Regression quantiles , author=. Econometrica , _pages=. 1978 , _publisher=

    1978

  9. [9]

    Proceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence , year=

    Variational inference for nonparametric Bayesian quantile regression , author=. Proceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence , year=

  10. [10]

    Statistics & Probability Letters , _volume=

    Bayesian quantile regression , author=. Statistics & Probability Letters , _volume=. 2001 , _publisher=

    2001

  11. [11]

    arXiv preprint arXiv:2407.12223 , year=

    Conditional Quantile Estimation for Uncertain Watch Time in Short-Video Recommendation , author=. arXiv preprint arXiv:2407.12223 , year=

    work page arXiv

  12. [12]

    Proceedings of the Nineteenth ACM Conference on Recommender Systems (RecSys '25) , pages=

    Multi-Granularity Distribution Modeling for Video Watch Time Prediction via Exponential-Gaussian Mixture Network , author=. Proceedings of the Nineteenth ACM Conference on Recommender Systems (RecSys '25) , pages=. 2025 , organization=

    2025

  13. [13]

    Proceedings of the 17th ACM Conference on Recommender Systems , pages=

    Uncovering user interest from biased and noised watch time in video recommendation , author=. Proceedings of the 17th ACM Conference on Recommender Systems , pages=

  14. [14]

    Sun, Jie and Ding, Zhaoying and Wang, Ben and others , booktitle=

  15. [15]

    Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    Tree based progressive regression model for watch-time prediction in short-video recommendation , author=. Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

  16. [16]

    Proceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining , pages=

    Deconfounding duration bias in watch-time prediction for video recommendation , author=. Proceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining , pages=

  17. [17]

    Proceedings of the 17th ACM conference on recommender systems , pages=

    Understanding and modeling passive-negative feedback for short-video sequential recommendation , author=. Proceedings of the 17th ACM conference on recommender systems , pages=

  18. [18]

    Proceedings of the 8th ACM Conference on Recommender systems , pages=

    Beyond clicks: dwell time for personalization , author=. Proceedings of the 8th ACM Conference on Recommender systems , pages=

  19. [19]

    ECAI 2020 , pages=

    Capturing attraction distribution: Sequential attentive network for dwell time prediction , author=. ECAI 2020 , pages=. 2020 , organization=

    2020

  20. [20]

    Fisheries Research , volume=

    Application of the Tweedie distribution to zero-catch data in CPUE analysis , author=. Fisheries Research , volume=. 2008 , publisher=

    2008

  21. [21]

    1994 , booktitle=

    Mixture density networks , author=. 1994 , booktitle=

    1994

  22. [22]

    Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD '18) , pages=

    Modeling task relationships in multi-task learning with multi-gate mixture-of-experts , author=. Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD '18) , pages=. 2018 , organization=

    2018

  23. [23]

    Proceedings of the 10th ACM conference on recommender systems , pages=

    Deep neural networks for youtube recommendations , author=. Proceedings of the 10th ACM conference on recommender systems , pages=. 2016 , organization=

    2016

  24. [24]

    Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    Deep interest network for click-through rate prediction , author=. Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , pages=. 2018 , organization=

    2018

  25. [25]

    Proceedings of the 31st ACM international conference on information & knowledge management , pages=

    Real-time short video recommendation on mobile devices , author=. Proceedings of the 31st ACM international conference on information & knowledge management , pages=. 2022 , organization=

    2022

  26. [26]

    Proceedings of the 31st ACM International Conference on Information & Knowledge Management (CIKM '22) , pages=

    KuaiRec: A Fully-Observed Dataset and Insights for Evaluating Recommender Systems , author=. Proceedings of the 31st ACM International Conference on Information & Knowledge Management (CIKM '22) , pages=. 2022 , organization=

    2022

  27. [27]

    Annual review of statistics and its application , volume=

    Finite mixture models , author=. Annual review of statistics and its application , volume=. 2019 , publisher=

    2019

  28. [28]

    Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    Counteracting Duration Bias in Video Recommendation via Counterfactual Watch Time , author=. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

  29. [29]

    Proceedings of the 16th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

    Overlapping experiment infrastructure: More, better, faster experimentation , author=. Proceedings of the 16th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

  30. [30]

    The annals of mathematical statistics , volume=

    On information and sufficiency , author=. The annals of mathematical statistics , volume=

  31. [31]

    Companion Proceedings of the ACM Web Conference 2023 , pages=

    Reinforcing user retention in a billion scale short video recommender system , author=. Companion Proceedings of the ACM Web Conference 2023 , pages=

    2023

  32. [32]

    Journal of applied statistics , volume=

    Beta regression for modelling rates and proportions , author=. Journal of applied statistics , volume=. 2004 , publisher=

    2004

  33. [33]

    What uncertainties do we need in

    Kendall, Alex and Gal, Yarin , booktitle=. What uncertainties do we need in

  34. [34]

    Advances in Neural Information Processing Systems (NeurIPS) , volume=

    Simple and scalable predictive uncertainty estimation using deep ensembles , author=. Advances in Neural Information Processing Systems (NeurIPS) , volume=

  35. [35]

    arXiv preprint arXiv:2006.07474 , year=

    Efficient epistemic uncertainty estimation in regression ensemble models using pairwise-distance estimators , author=. arXiv preprint arXiv:2006.07474 , year=

    work page arXiv 2006

  36. [36]

    Advances in Neural Information Processing Systems (NeurIPS) , volume=

    Deep evidential regression , author=. Advances in Neural Information Processing Systems (NeurIPS) , volume=

  37. [37]

    Advances in Neural Information Processing Systems (NeurIPS) , volume=

    Simple and principled uncertainty estimation with deterministic deep learning via distance awareness , author=. Advances in Neural Information Processing Systems (NeurIPS) , volume=

  38. [38]

    Advances in Neural Information Processing Systems (NeurIPS) , volume=

    Conformalized quantile regression , author=. Advances in Neural Information Processing Systems (NeurIPS) , volume=

  39. [39]

    Proceedings of the 37th Conference on Uncertainty in Artificial Intelligence (UAI) , pages=

    Integrating uncertainty awareness into conformalized quantile regression , author=. Proceedings of the 37th Conference on Uncertainty in Artificial Intelligence (UAI) , pages=

  40. [40]

    International Conference on Machine Learning (ICML) , pages=

    A distributional perspective on reinforcement learning , author=. International Conference on Machine Learning (ICML) , pages=

  41. [41]

    Variational

    Harrison, James and Willes, John and Snoek, Jasper , booktitle=. Variational

  42. [42]

    Biometrika , volume=

    On optimal and data-based histograms , author=. Biometrika , volume=. 1979 , publisher=

    1979

  43. [43]

    and Ruan, T

    Lian, C. and Ruan, T. and Denoeux, T. , journal=

  44. [44]

    International Conference on Learning Representations (ICLR) , year=

    Spectral normalization for generative adversarial networks , author=. International Conference on Learning Representations (ICLR) , year=

  45. [45]

    International Conference on Machine Learning (ICML) , pages=

    Dropout as a bayesian approximation: Representing model uncertainty in deep learning , author=. International Conference on Machine Learning (ICML) , pages=. 2016 , organization=

    2016

  46. [46]

    Springer Science & Business Media , year=

    Algorithmic learning in a random world , author=. Springer Science & Business Media , year=

  47. [47]

    The Annals of Statistics , volume=

    Predictive inference with the jackknife+ , author=. The Annals of Statistics , volume=. 2021 , publisher=

    2021

  48. [48]

    Springer , year=

    Pattern recognition and machine learning , author=. Springer , year=

  49. [49]

    International Conference on Machine Learning (ICML) , pages=

    Probabilistic backpropagation for scalable learning of bayesian neural networks , author=. International Conference on Machine Learning (ICML) , pages=. 2015 , organization=

    2015

  50. [50]

    Advances in Neural Information Processing Systems (NeurIPS) , volume=

    Single-model uncertainties for deep learning , author=. Advances in Neural Information Processing Systems (NeurIPS) , volume=

  51. [51]

    International Conference on Machine Learning (ICML) , pages=

    On calibration of modern neural networks , author=. International Conference on Machine Learning (ICML) , pages=. 2017 , organization=

    2017

  52. [52]

    International Conference on Machine Learning (ICML) , pages=

    Uncertainty estimation using a single deep deterministic neural network , author=. International Conference on Machine Learning (ICML) , pages=. 2020 , organization=

    2020

  53. [53]

    Biometrika , volume=

    On the likelihood that one unknown probability exceeds another in view of the evidence of two samples , author=. Biometrika , volume=

  54. [54]

    Foundations and Trends® in Machine Learning , volume=

    A tutorial on thompson sampling , author=. Foundations and Trends® in Machine Learning , volume=

  55. [55]

    Advances in Neural Information Processing Systems (NeurIPS) , volume=

    Variational Dropout and the Local Reparameterization Trick , author=. Advances in Neural Information Processing Systems (NeurIPS) , volume=

  56. [56]

    NeurIPS , year=

    What uncertainties do we need in bayesian deep learning for computer vision? , author=. NeurIPS , year=

  57. [57]

    2018 , publisher=

    Foundations of machine learning , author=. 2018 , publisher=

    2018

  58. [58]

    Engineering , volume=

    Non-IID recommender systems: A review and framework of recommendation paradigm shifting , author=. Engineering , volume=. 2016 , note=

    2016

  59. [59]

    ICML , year=

    Weight uncertainty in neural network , author=. ICML , year=

  60. [60]

    Journal of the American Statistical Association , volume=

    Strictly proper scoring rules, prediction, and estimation , author=. Journal of the American Statistical Association , volume=. 2007 , publisher=

    2007

  61. [61]

    Journal of Machine Learning Research , volume=

    Nonparametric quantile estimation , author=. Journal of Machine Learning Research , volume=

  62. [62]

    Soviet Mathematics , volume=

    Estimation of an unknown distribution density from observations , author=. Soviet Mathematics , volume=

  63. [63]

    Bernoulli , volume=

    Estimating conditional quantiles with the help of the pinball loss , author=. Bernoulli , volume=. 2011 , publisher=

    2011

  64. [64]

    Proceedings of the 35th International Conference on Machine Learning , pages=

    Decomposition of Uncertainty in Bayesian Deep Learning for Efficient and Risk-sensitive Learning , author=. Proceedings of the 35th International Conference on Machine Learning , pages=. 2018 , volume=

    2018

  65. [65]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    Learning for Single-Shot Confidence Calibration in Deep Neural Networks through Stochastic Inferences , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

  66. [66]

    Kristiadi, Agustinus and Hein, Matthias and Hennig, Philipp , booktitle=. Being. 2020 , publisher=

    2020

  67. [67]

    Tashiro, Yusuke and Song, Jiaming and Song, Yang and Ermon, Stefano , booktitle=

  68. [68]

    2018 IEEE International Conference on Robotics and Automation (ICRA) , pages=

    End-to-end driving via conditional imitation learning , author=. 2018 IEEE International Conference on Robotics and Automation (ICRA) , pages=. 2018 , organization=

    2018

  69. [69]

    , booktitle=

    Yang, Zhilin and Dai, Zihang and Salakhutdinov, Ruslan and Cohen, William W. , booktitle=. Breaking the Softmax Bottleneck: A High-Rank. 2018 , url=

    2018

  70. [70]

    2000 , publisher=

    Asymptotic statistics , author=. 2000 , publisher=

    2000

  71. [71]

    2019 International Conference on Robotics and Automation (ICRA) , pages=

    Multimodal trajectory predictions for autonomous driving using deep convolutional networks , author=. 2019 International Conference on Robotics and Automation (ICRA) , pages=. 2019 , organization=

    2019

  72. [72]

    Bayesian Active Learning for Classification and Preference Learning

    Bayesian active learning for classification and preference learning , author=. arXiv preprint arXiv:1112.5745 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  73. [73]

    Proceedings of the 34th International Conference on Machine Learning (ICML) , pages=

    Deep bayesian active learning with image data , author=. Proceedings of the 34th International Conference on Machine Learning (ICML) , pages=. 2017 , organization=

    2017

  74. [74]

    Advances in Neural Information Processing Systems (NeurIPS) , volume=

    BatchBALD: Efficient and diverse batch active learning for deep neural networks , author=. Advances in Neural Information Processing Systems (NeurIPS) , volume=

  75. [75]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    Learning loss for active learning , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

  76. [76]

    International Conference on Learning Representations (ICLR) , year=

    Active Learning for Convolutional Neural Networks: A Core-Set Approach , author=. International Conference on Learning Representations (ICLR) , year=