Wheat physiology predictor: predicting physiological traits in wheat from hyperspectral reflectance measurements using deep learning

Table 3 R² for performance of various models on the test set

Model	LMA	N_area	SPAD	N_mass	V_cmax	V_cmax25	J	A	g_s	V_cmax25/N_area	Mean
Multitask 1DCNN*	0.867	0.931	0.833	0.807	0.781	0.779	0.858	0.730	0.502	0.480	0.757
Multitask 1DCNN, No Spectral Trim	0.887	0.946	0.857	0.855	0.776	0.770	0.863	0.762	0.526	0.521	0.776
Singletask 1DCNN*	0.855	0.955	0.866	0.715	0.796	0.740	0.846	0.689	0.496	0.444	0.740
Singletask 1DCNN, No Spectral Trim	0.880	0.964	0.860	0.799	0.809	0.763	0.849	0.710	0.499	0.493	0.763
MLP	0.856	0.912	0.862	0.713	0.752	0.721	0.791	0.672	0.423	0.492	0.719
LSTM	0.704	0.809	0.748	0.685	0.578	0.640	0.732	0.615	0.352	-0.067	0.580
PLSR*	0.885	0.944	0.841	0.789	0.770	0.656	0.853	0.667	0.427	0.576	0.741
XGBoost	0.797	0.941	0.824	0.734	0.632	0.674	0.775	0.657	0.353	0.234	0.662
Ensemble	0.895	0.959	0.866	0.832	0.822	0.762	0.876	0.751	0.521	0.579	0.785

The results reported are the mean of three runs with different random seeds. Those marked with an asterisk are included in the ensemble model

ISSN: 1746-4811