Uncertainty quantification in Neural Networks by Approximate Bayesian Computation: Application to fatigue in composite materials

Fernández, Juan; Chiachío, Manuel; Chiachío, Juan; Muñoz, Rafael; Herrera, Francisco

doi:10.1016/j.engappai.2021.104511

Cited by 34 publications

(6 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The performance of the Bayesian algorithms described in Section 3.2, evaluated on the CFRP Composites Dataset from NASA Ames Prognostics Data Repository, was discussed in Table 1 of (Fernández et al, 2022), where the The probability of failure has been calculated for the last cycles of the experiment, following the methodology explained in Section 3.3, and the results are shown in Table 1. It can be seen that BNN by ABC-SS provides the closest probabilities to the observed data.…”

Section: Resultsmentioning

confidence: 99%

“…, j. The interested reader is referred to (Chiachio et al, 2014) for further information about ABC-SS, and to (Fernández et al, 2022) for details about the implementation of BNN by ABC-SS.…”

Section: Bnn By Abc-ssmentioning

confidence: 99%

“…• BNN by ABC-SS: A BNN trained with Algorithm 1 of (Fernández et al, 2022), adapted with a while loop and σ j = σ 0 p. The hyper-parameters used are P 0 =0.1, N =100,000, σ 0 =0.75, p=0.58 and tolerance value ϵ=0.025. The value of σ and p are chosen by a trial and error process, very much like the learning rate in standard backpropagation.…”

Section: Baseline Algorithms and Metricsmentioning

confidence: 99%

“…All that translates into a rigid quantification of the uncertainty, and a certain predisposition to problems such as instability or exploding/vanishing gradient. Contrariwise, BNN by ABC-SS (Fernández, Chiachío, Chiachío, Muñoz, & Herrera, 2022) has proven great flexibility to capture the uncertainty inherent in the observed data, thanks to its gradient-free nature, the non-parametric formulation of the weights and the absence of likelihood/cost function.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Probabilistic Safety Assessment in Composite Materials using BNN by ABC-SS

Fernández

Chiachío²,

Chiachío³

et al. 2022

PHM_CONF

View full text Add to dashboard Cite

Carbon fiber reinforced polymer composites present excellent mechanical properties, however, their behaviour under fatigue and the interaction between the different failure modes is not yet well understood. This uncertainty, or lack of knowledge, is the reason why they are still not extensively used in the aerospace industry, where safety is critical. In this paper, Bayesian neural networks trained with approximate Bayesian computation (BNN by ABC-SS) are used to quantify such uncertainty and undertake a probabilistic safety assessment. An experiment is carried out using data from composite fatigue testing, where the proposed algorithm is compared against the state-of-the-art Bayesian neural networks. The results show that, the flexibility of BNN by ABC-SS to quantify the uncertainty significantly contributes towards a reliable safety assessment. Measuring the unknowns with confidence can be crucial when safety is at stake.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Bnn By Abc-ssmentioning

confidence: 99%

Section: Baseline Algorithms and Metricsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Probabilistic Safety Assessment in Composite Materials using BNN by ABC-SS

Fernández

Chiachío²,

Chiachío³

et al. 2022

PHM_CONF

View full text Add to dashboard Cite

show abstract

“…Machine learning is one of the effective technique for classifying ASD using the behaviour of brain and features of brain, like motor activities, body gestures, facial expressions, and eye contact (Xie et al, 2019). In (Heinsfeld et al, 2018), deep learning approach (Odusami et al, 2021;RukhMuzammil et al, 2021) is employed for classifying the ASD disorders in which the deep learning technique is designed by the integration of supervised as well as unsupervised machine learning approaches but it has the problem of uncertainty quantification (Abdar, Fahami, et al, 2021;Abdar, Pourpanah, et al, 2021;Abdar, Salari, et al, 2021;Abdar, Samami, et al, 2021;Fernández et al, 2022;Qin et al, 2021). The deep learning approaches can automatically extracts the features using data-assisted learning process without the manual assistance for further effective classification.…”

mentioning

confidence: 99%

Feedback Artificial Tree‐Anti Corona Virus Optimization enabled deep learning for detecting autism spectrum disorder

Arumugam¹,

Velrajkumar

Sriramakrishnan

et al. 2022

Expert Systems

View full text Add to dashboard Cite

Autism spectrum disorder (ASD) is an umbrella term for a number of neurodevelopmental conditions with many heterogeneous behavioural indications. Recent medical imaging approaches use functional Magnetic Resonance Imaging (fMRI) for human recognition of the various neurological syndromes. However, these traditional techniques are time consuming and expensive. Thus, in this research, an optimization assisted deep learning technique, named Feedback Artificial Virus Optimization (FAVO)‐based deep residual network (DRN), is developed. FAVO‐based DRN is designed to incorporate the Feedback Artificial Tree (FAT) algorithm with Anti Corona Virus Optimization (ACVO). First, Region‐Of‐Interest extraction is carried out using thresholding techniques with nub region extraction completed using the proposed FAVO algorithm. ASD classification is then carried out using a DRN classifier. Evaluation of the proposal uses the ABIDE‐1 and ABIDE‐2 datasets. The developed FAVO algorithm attains better accuracy, sensitivity, and specificity of 0.9214, 0.9365, and 0.9142, respectively, by considering ABIDE‐2 dataset.

show abstract