Uncertainty quantification of spectral predictions using deep neural networks
Sneha Verma, Nik Khadijah Nik Aznan, Kathryn Garside, Thomas J. Penfold
Abstract
We demonstrate uncertainty quantification for deep neural network predictions of transition metal X-ray absorption near-edge structure spectra. Our results not only provide accurate spectral predictions, but reliably assess when the model fails.
Topics & Concepts
Artificial neural networkDeep neural networksSpectral lineEnhanced Data Rates for GSM EvolutionComputer scienceArtificial intelligenceAbsorption (acoustics)Biological systemPattern recognition (psychology)Statistical physicsPhysicsOpticsAstronomyBiologyMachine Learning in Materials ScienceElectron and X-Ray Spectroscopy TechniquesAdvanced X-ray and CT Imaging