Uncertainty Quantification for Deep Learning Prognostics

Assessment of the latest uncertainty quantification (UQ) methods for prognostics in DL

The review is done with the following paper:
Luis Basora, , Arthur Viens, Manuel Arias Chao, and Xavier Olive. “A Benchmark on Uncertainty Quantification for Deep Learning Prognostics.” (2023).

A Benchmark on Uncertainty Quantification for Deep Learning Prognostics

Table of Contents

  1. Abstract of the paper
  2. Bayesian Neural Network
  3. Backpropagation Methods
  4. Dataset Explanation
  5. Model Architecture
  6. Training Phase
  7. Evalution and Results
  8. Conclusion

Abstract of the paper

The following paper contains the assessment of the latest uncertainty quantification (UQ) methods for prognostics in deep learning (DL).

  • UQ assessment: remaining useful lifetime (RUL) prediction
    • NASA’s N-CMAPSS dataset
  • Methods or techniques will be used for the assessment:
    • Bayesian Neural Network (BNN) … will be focusing on BNN only
    • Monte Carlo Dropout (MCD)
    • Heteroscedastic Neural Network (HNN)
    • Deep Ensemble (DE)

Bayesian Neural Network

Bayes Theorem

see Bayes’ Theorem.

\[\begin{align} \text{Posterior} \propto \text{Prior} \times \text{Likelihood} \\ P(A|B) = \frac{P(B|A)\times P(A)}{P(B)} \\ P(w|\mathcal{D}) \propto P(\mathcal{D}|w) \times P(w) \\ \end{align}\]

Variational Inference

\[P(w|\mathcal{D}) \propto P(\mathcal{D}|w) \times P(w)\]

Kullback-Leibler Divergence

Bayes by Backprop.

Backpropagation Methods

Dataset Explanation

Model Architecture

Training Phase

Evaluation and Results

Conclusion

References

2023

  1. A Benchmark on Uncertainty Quantification for Deep Learning Prognostics
    Luis Basora, Arthur Viens, Manuel Arias Chao, and 1 more author
    2023