Towards calibrated and scalable uncertainty representations for neural networks

Seedat, Nabeel; Kanan, Christopher

doi:10.48550/arxiv.1911.00104

Cited by 4 publications

(3 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Deep neural networks are deterministic in nature, which is not always desirable, especially in the case of high-risk systems such as clinical systems [85,86,87]. It is desirable to have a system that can provide associated confidence with its prediction.…”

Section: Monte Carlo (Mc) Dropoutmentioning

confidence: 99%

SACDNet: Towards Early Type 2 Diabetes Prediction with Uncertainty for Electronic Health Records

Nasir¹,

Malik²

2023

Preprint

View full text Add to dashboard Cite

show abstract

Section: Monte Carlo (Mc) Dropoutmentioning

confidence: 99%

SACDNet: Towards Early Type 2 Diabetes Prediction with Uncertainty for Electronic Health Records

Nasir¹,

Malik²

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Benchmarks are an important tool to help researchers prioritize the right approaches and to inform practitioners which methods are suited for their applications [5]. There is a growing demand for benchmarking in BDL, since methods must be scored both for task performance and uncertainty quality [6,7]. Rigorously evaluating the latter is considerably more difficult, since depending on the problem setting no direct uncertainty ground-truth exists, requiring a well-defined experimental setup [8].…”

Section: A Related Workmentioning

confidence: 99%

Benchmarking Scalable Predictive Uncertainty in Text Classification

Landeghem

Blaschko

Anckaert³

et al. 2022

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Uncertainty correction (EpiCC) [59] Variational Inference (Stochastic Variational Inference) Initial adaptation [60], Bayes by Backprop [61], Reparameterization Trick [62] Normalizing flows [63][64] (Monte Carlo Dropout) MC-Dropout [65], MC-DropConnect [54], Heteroscedastic classification NN [14] Figure 11) Uncertainty Quantification techniques for deep learning models Sampling Methods (Markov Chain Monte Carlo) Hamiltonian Monte Carlo [66], Stochastic Gradient Monte Carlo (SG-MCMC) [67], RECAST (SG-MCMC method) [68] Laplace Approx.…”

Section: Standard Bnnsmentioning

confidence: 99%

Machine Learning & Uncertainty Quantification: Application in Building Energy Consumption

Fakour

Parhizkar

Ramezani

et al. 2022

2022 Annual Reliability and Maintainability Symposium (RAMS)

View full text Add to dashboard Cite

The adaptation and use of Machine Learning (ML) in our daily lives has led to concerns in lack of transparency, privacy, reliability, among others. As a result, we are seeing research in niche areas such as interpretability, causality, bias and fairness, and reliability. In this survey paper, we focus on a critical concern for adaptation of ML in risksensitive applications, namely understanding and quantifying uncertainty. Our paper approaches this topic in a structured way, providing a review of the literature in the various facets that uncertainty is enveloped in the ML process. We begin by defining uncertainty and its categories (e.g., aleatoric and epistemic), understanding sources of uncertainty (e.g., data and model), and how uncertainty can be assessed in terms of uncertainty quantification techniques (Ensembles, Bayesian Neural Networks, etc.). As part of our assessment and understanding of uncertainty in the ML realm, we cover metrics for uncertainty quantification for a single sample, dataset, and metrics for accuracy of the uncertainty estimation itself. This is followed by discussions on calibration (model and uncertainty), and decision making under uncertainty. Thus, we provide a more complete treatment of uncertainty: from the sources of uncertainty to the decision-making process. We have focused the review of uncertainty quantification methods on Deep Learning (DL), while providing the necessary background for uncertainty discussion within ML in general. Key contributions in this review are broadening the scope of uncertainty discussion, as well as an updated review of uncertainty quantification methods in DL.

show abstract

Towards calibrated and scalable uncertainty representations for neural networks

Cited by 4 publications

References 7 publications

SACDNet: Towards Early Type 2 Diabetes Prediction with Uncertainty for Electronic Health Records

SACDNet: Towards Early Type 2 Diabetes Prediction with Uncertainty for Electronic Health Records

Benchmarking Scalable Predictive Uncertainty in Text Classification

Machine Learning & Uncertainty Quantification: Application in Building Energy Consumption

Contact Info

Product

Resources

About