Uncertainty quantification of spectral predictions using deep neural networks

Verma, Sneha; Aznan, Nik Khadijah Nik; Garside, Kathryn; Penfold, Thomas J.

doi:10.1039/d3cc01988h

Cited by 5 publications

(15 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…52 Recently work addressing the forward mapping ( i.e. structure to spectra), highlighted limitations of wACSF descriptor for capturing the focusing effect giving rise to large uncertainties in the associated predictions, 32 and our present results clearly exhibit similar limitations for the reverse spectrum to spectra mapping. Fe(dedtc) 3 also gives a poor prediction but in contrast to the previous examples a low uncertainty, with the model exhibiting a distinct over-confidence.…”

Section: Resultssupporting

confidence: 69%

“…8b, appear at larger MSE. This suggests, in contrast to recent work 32,36 that when the model uncertainty fails, it is slightly over-confident, i.e. the MSE is larger than suggested from that specific σ , which is discussed in the next section.…”

Section: Resultscontrasting

confidence: 64%

“…8a shows a histogram of the coverage, defined as the percentage of target data points which fall within ±2 σ of the average prediction computed using the bootstrap resampling approach for the held-out set. This demonstrates a distribution largely between 45–100%, with a median coverage of 64%, which is smaller than observed for the forward structure to spectrum network, 32 but comparable to that reported for X-ray emission spectroscopy. 14 While the coverage demonstrates the performance of the uncertainty quantification methods, it can only be used as a metric if the known structure and therefore target G 2 wACSF exists, which would negate the purpose for the use of DNN for future applications.…”

Section: Resultssupporting

confidence: 59%

“…Having established the performance of the network in terms of predicting the G 2 wACSF, we now seek to assess uncertainty using the bootstrap resampling approach outlined in ref. 32. Fig.…”

Section: Estimating Uncertainty Of the Structural Predictionsmentioning

confidence: 99%

“…We have recently demonstrated that the bootstrap resampling technique can be used to estimate the uncertainty arising from neural network predictions of XANES spectral shapes from input structures. 32 Here, we adopt the same approach estimating the uncertainty in the estimated structural predictions. Briefly, N machine learning models are optimised using N reference datasets sampled with replacement from the original reference dataset; each one of these is the same size as the original reference dataset and, consequently, may contain repeated instances of the same sample.…”

Section: Computational and Technical Detailsmentioning

confidence: 99%

See 4 more Smart Citations

Towards the automated extraction of structural information from X-ray absorption spectra

David,

Nik Aznan,

Garside

et al. 2023

Digital Discovery

Self Cite

View full text Add to dashboard Cite

show abstract

Section: Resultssupporting

confidence: 69%

Section: Resultscontrasting

confidence: 64%

Section: Resultssupporting

confidence: 59%

“…Having established the performance of the network in terms of predicting the G 2 wACSF, we now seek to assess uncertainty using the bootstrap resampling approach outlined in ref. 32. Fig.…”

Section: Estimating Uncertainty Of the Structural Predictionsmentioning

confidence: 99%

Section: Computational and Technical Detailsmentioning

confidence: 99%

See 3 more Smart Citations

Towards the automated extraction of structural information from X-ray absorption spectra

David,

Nik Aznan,

Garside

et al. 2023

Digital Discovery

Self Cite

View full text Add to dashboard Cite

show abstract

Uncertainty Qualification for Deep Learning-Based Elementary Reaction Property Prediction

Liu,

Mo,

Cheng

2024

J. Chem. Inf. Model.

View full text Add to dashboard Cite

The prediction of the thermodynamic and kinetic properties of elementary reactions has shown rapid improvement due to the implementation of deep learning (DL) methods. While various studies have reported the success in predicting reaction properties, the quantification of prediction uncertainty has seldom been investigated, thus compromising the confidence in using these predicted properties in practical applications. Here, we integrated graph convolutional neural networks (GCNN) with three uncertainty prediction techniques, including deep ensemble, Monte Carlo (MC)-dropout, and evidential learning, to provide insights into the uncertainty quantification and utility. The deep ensemble model outperforms others in accuracy and shows the highest reliability in estimating prediction uncertainty across all elementary reaction property data sets. We also verified that the deep ensemble model showed a satisfactory capability in recognizing epistemic and aleatoric uncertainties. Additionally, we adopted a Monte Carlo Tree Search method for extracting the explainable reaction substructures, providing a chemical explanation for DL predicted properties and corresponding uncertainties. Finally, to demonstrate the utility of uncertainty qualification in practical applications, we performed an uncertainty-guided calibration of the DL-constructed kinetic model, which achieved a 25% higher hit ratio in identifying dominant reaction pathways compared to that of the calibration without uncertainty guidance.

show abstract

Artificial-Intelligence-Enhanced On-the-Fly Simulation of Nonlinear Time-Resolved Spectra

Pios,

Gelin,

Ullah

et al. 2024

J. Phys. Chem. Lett.

View full text Add to dashboard Cite

Time-resolved spectroscopy is an important tool for unraveling the minute details of structural changes in molecules of biological and technological significance. The nonlinear femtosecond signals detected for such systems must be interpreted, but it is a challenging task for which theoretical simulations are often indispensable. Accurate simulations of transient absorption or two-dimensional electronic spectra are, however, computationally very expensive, prohibiting the wider adoption of existing first-principles methods. Here, we report an artificial-intelligence-enhanced protocol to drastically reduce the computational cost of simulating nonlinear time-resolved electronic spectra, which makes such simulations affordable for polyatomic molecules of increasing size. The protocol is based on the doorway−window approach for the on-the-fly surface-hopping simulations. We show its applicability for the prototypical molecule of pyrazine for which it produces spectra with high precision with respect to ab initio reference while cutting the computational cost by at least 95% compared to pure first-principles simulations.

show abstract

Uncertainty quantification of spectral predictions using deep neural networks

Abstract: We investigate the performance of uncertainty quantification methods, namely deep ensembles and bootstrap resampling, for deep neural network (DNN) predictions of transition metal K-edge X-ray absorption near-edge structure (XANES) spectra....

Cited by 5 publications

References 18 publications

Towards the automated extraction of structural information from X-ray absorption spectra

Towards the automated extraction of structural information from X-ray absorption spectra

Uncertainty Qualification for Deep Learning-Based Elementary Reaction Property Prediction

Artificial-Intelligence-Enhanced On-the-Fly Simulation of Nonlinear Time-Resolved Spectra

Contact Info

Product

Resources

About