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. Bootstrap resampling combined with our multi-layer perceptron (MLP) model provides an accurate assessment of uncertainty with >90% of all predicted spectral intensities falling within ±3σ of the true values for held-out data across the nine first-row transition metal K-edge XANES spectra.