-
A Benchmark on Uncertainty Quantification for Deep Learning Prognostics
Authors:
Luis Basora,
Arthur Viens,
Manuel Arias Chao,
Xavier Olive
Abstract:
Reliable uncertainty quantification on RUL prediction is crucial for informative decision-making in predictive maintenance. In this context, we assess some of the latest developments in the field of uncertainty quantification for prognostics deep learning. This includes the state-of-the-art variational inference algorithms for Bayesian neural networks (BNN) as well as popular alternatives such as…
▽ More
Reliable uncertainty quantification on RUL prediction is crucial for informative decision-making in predictive maintenance. In this context, we assess some of the latest developments in the field of uncertainty quantification for prognostics deep learning. This includes the state-of-the-art variational inference algorithms for Bayesian neural networks (BNN) as well as popular alternatives such as Monte Carlo Dropout (MCD), deep ensembles (DE) and heteroscedastic neural networks (HNN). All the inference techniques share the same inception deep learning architecture as a functional model. We performed hyperparameter search to optimize the main variational and learning parameters of the algorithms. The performance of the methods is evaluated on a subset of the large NASA NCMAPSS dataset for aircraft engines. The assessment includes RUL prediction accuracy, the quality of predictive uncertainty, and the possibility to break down the total predictive uncertainty into its aleatoric and epistemic parts. The results show no method clearly outperforms the others in all the situations. Although all methods are close in terms of accuracy, we find differences in the way they estimate uncertainty. Thus, DE and MCD generally provide more conservative predictive uncertainty than BNN. Surprisingly, HNN can achieve strong results without the added training complexity and extra parameters of the BNN. For tasks like active learning where a separation of epistemic and aleatoric uncertainty is required, radial BNN and MCD seem the best options.
△ Less
Submitted 9 February, 2023;
originally announced February 2023.
-
Integral Ricci curvature and the mass gap of Dirichlet Laplacians on domains
Authors:
Xavier Ramos Olivé,
Christian Rose,
Lili Wang,
Guofang Wei
Abstract:
We obtain a fundamental gap estimate for classes of bounded domains with quantitative control on the boundary in a complete manifold with integral bounds on the negative part of the Ricci curvature. This extends the result of \cite{Oden-Sung-Wang99} to $L^p$-Ricci curvature assumptions, $p>n/2$. To achieve our result, it is shown that the domains under consideration are John domains, what enables…
▽ More
We obtain a fundamental gap estimate for classes of bounded domains with quantitative control on the boundary in a complete manifold with integral bounds on the negative part of the Ricci curvature. This extends the result of \cite{Oden-Sung-Wang99} to $L^p$-Ricci curvature assumptions, $p>n/2$. To achieve our result, it is shown that the domains under consideration are John domains, what enables us to obtain an estimate on the first nonzero Neumann eigenvalue, which is of independent interest.
△ Less
Submitted 23 September, 2021;
originally announced September 2021.
-
Manifolds with bounded integral curvature and no positive eigenvalue lower bounds
Authors:
Connor C. Anderson,
Xavier Ramos Olivé,
Kamryn Spinelli
Abstract:
We provide an explicit construction of a sequence of closed surfaces with uniform bounds on the diameter and on $L^p$ norms of the curvature, but without a positive lower bound on the first non-zero eigenvalue of the Laplacian $λ_1$. This example shows that the assumption of smallness of the $L^p$ norm of the curvature is a necessary condition to derive Lichnerowicz and Zhong-Yang type estimates u…
▽ More
We provide an explicit construction of a sequence of closed surfaces with uniform bounds on the diameter and on $L^p$ norms of the curvature, but without a positive lower bound on the first non-zero eigenvalue of the Laplacian $λ_1$. This example shows that the assumption of smallness of the $L^p$ norm of the curvature is a necessary condition to derive Lichnerowicz and Zhong-Yang type estimates under integral curvature conditions.
△ Less
Submitted 2 November, 2021; v1 submitted 22 March, 2021;
originally announced March 2021.
-
Quantitative Sobolev extensions and the Neumann heat kernel for integral Ricci curvature conditions
Authors:
Olaf Post,
Xavier Ramos Olivé,
Christian Rose
Abstract:
We prove the existence of Sobolev extension operators for certain uniform classes of domains in a Riemannian manifold with an explicit uniform bound on the norm depending only on the geometry near their boundaries. We use this quantitative estimate to obtain uniform Neumann heat kernel upper bounds and gradient estimates for positive solutions of the Neumann heat equation assuming integral Ricci c…
▽ More
We prove the existence of Sobolev extension operators for certain uniform classes of domains in a Riemannian manifold with an explicit uniform bound on the norm depending only on the geometry near their boundaries. We use this quantitative estimate to obtain uniform Neumann heat kernel upper bounds and gradient estimates for positive solutions of the Neumann heat equation assuming integral Ricci curvature conditions and geometric conditions on the boundary. Those estimates also imply quantitative lower bounds on the first Neumann eigenvalue of the considered domains.
△ Less
Submitted 8 July, 2020;
originally announced July 2020.
-
Quantitative Assessments of Runway Excursion Precursors using Mode S data
Authors:
Xavier Olive,
Pierre Bieber
Abstract:
A way to assess rare aircraft incidents (e.g., runway excursion) is to identify contributing factors (e.g., late braking, long landing, inappropriate flare, unstable approach) and to build a dependency tree (e.g., long landing may be the result of an unstable approach not followed by a go around) that describes the causality between these factors. Probabilities are then fed into such models in ord…
▽ More
A way to assess rare aircraft incidents (e.g., runway excursion) is to identify contributing factors (e.g., late braking, long landing, inappropriate flare, unstable approach) and to build a dependency tree (e.g., long landing may be the result of an unstable approach not followed by a go around) that describes the causality between these factors. Probabilities are then fed into such models in order to evaluate the assessed risk. When estimating such probabilities, many sources can be of interest. Airlines have access to the comprehensive flight data records of their fleet; manufacturers push to collect data for the aircraft they build; air traffic control log radar tracks. Albeit not as complete as other flight data records, Mode S data is very attractive, esp. for academics, as the data is open, may be published without obfuscation and offers reproducible results to the community. Mode S also provides an indiscriminate source of information (not limited to an airline or to an aircraft type) that is of great help for putting in context flights matching unusual patterns. We propose to discuss the advantages and limitations of an analysis based only on Mode S data with a case study around the runway excursion risk assessment.
△ Less
Submitted 27 March, 2019;
originally announced March 2019.
-
Zhong-Yang type eigenvalue estimate with integral curvature condition
Authors:
Xavier Ramos Olivé,
Shoo Seto,
Guofang Wei,
Qi S. Zhang
Abstract:
We prove a sharp Zhong-Yang type eigenvalue lower bound for closed Riemannian manifolds with control on integral Ricci curvature.
We prove a sharp Zhong-Yang type eigenvalue lower bound for closed Riemannian manifolds with control on integral Ricci curvature.
△ Less
Submitted 24 March, 2019; v1 submitted 3 December, 2018;
originally announced December 2018.
-
French SKA White Book - The French Community towards the Square Kilometre Array
Authors:
F. Acero,
J. -T. Acquaviva,
R. Adam,
N. Aghanim,
M. Allen,
M. Alves,
R. Ammanouil,
R. Ansari,
A. Araudo,
E. Armengaud,
B. Ascaso,
E. Athanassoula,
D. Aubert,
S. Babak,
A. Bacmann,
A. Banday,
K. Barriere,
F. Bellossi,
J. -P. Bernard,
M. G. Bernardini,
M. Béthermin,
E. Blanc,
L. Blanchet,
J. Bobin,
S. Boissier
, et al. (153 additional authors not shown)
Abstract:
The "Square Kilometre Array" (SKA) is a large international radio telescope project characterised, as suggested by its name, by a total collecting area of approximately one square kilometre, and consisting of several interferometric arrays to observe at metric and centimetric wavelengths. The deployment of the SKA will take place in two sites, in South Africa and Australia, and in two successive p…
▽ More
The "Square Kilometre Array" (SKA) is a large international radio telescope project characterised, as suggested by its name, by a total collecting area of approximately one square kilometre, and consisting of several interferometric arrays to observe at metric and centimetric wavelengths. The deployment of the SKA will take place in two sites, in South Africa and Australia, and in two successive phases. From its Phase 1, the SKA will be one of the most formidable scientific machines ever deployed by mankind, and by far the most impressive in terms of data throughput and required computing power. With the participation of almost 200 authors from forty research institutes and six private companies, the publication of this French SKA white paper illustrates the strong involvement in the SKA project of the French astronomical community and of a rapidly growing number of major scientific and technological players in the fields of Big Data, high performance computing, energy production and storage, as well as system integration.
△ Less
Submitted 28 March, 2018; v1 submitted 19 December, 2017;
originally announced December 2017.
-
Neumann Li-Yau gradient estimate under integral Ricci curvature bounds
Authors:
Xavier Ramos Olivé
Abstract:
We prove a Li-Yau gradient estimate for positive solutions to the heat equation, with Neumann boundary conditions, on a compact Riemannian submanifold with boundary ${\bf M}^n\subseteq {\bf N}^n$, satisfying the integral Ricci curvature assumption: \begin{equation} D^2 \sup_{x\in {\bf N}} \left( \oint_{B(x,D)} |Ric^-|^p dy \right)^{\frac{1}{p}} < K \end{equation} for $K(n,p)$ small enough,…
▽ More
We prove a Li-Yau gradient estimate for positive solutions to the heat equation, with Neumann boundary conditions, on a compact Riemannian submanifold with boundary ${\bf M}^n\subseteq {\bf N}^n$, satisfying the integral Ricci curvature assumption: \begin{equation} D^2 \sup_{x\in {\bf N}} \left( \oint_{B(x,D)} |Ric^-|^p dy \right)^{\frac{1}{p}} < K \end{equation} for $K(n,p)$ small enough, $p>n/2$, where $diam({\bf M})\leq D$. The boundary of ${\bf M}$ is not necessarily convex, but it needs to satisfy the interior rolling $R-$ball condition.
△ Less
Submitted 12 April, 2018; v1 submitted 24 October, 2017;
originally announced October 2017.
