Fig. 1: DLRN work-flow analysis of 2D images.
From: Deep Learning Reaction Framework (DLRN) for kinetic modeling of time-resolved data
a Example of 2D spectra data used for the analysis. b One-hot encoding prediction obtained from DLRN using the model block. Each position represents a specific kinetic model. c Conversion of the one-hot encoding prediction into a model matrix. This model matrix is used in both time and amplitude blocks for predicting time constants and amplitudes. d Time constant predictions obtained from DLRN using the time block. e Normalized amplitude prediction obtained from DLRN using the amplitude block. f Comparison between expected and predicted models with the corresponding time constant for each decay pathway. g Comparison between expected and predicted spectra. The predictions of model, time constants, and amplitudes match well the expected results. h Accuracy values for model, tau and amplitude prediction obtained using DLRN on 100k unknown 2D datasets (See also Table SI 1).
