• Spectroscopy and Spectral Analysis
  • Vol. 42, Issue 11, 3647 (2022)
Schematic diagram of analysis of different spectral bands and resolution
Fig. 1. Schematic diagram of analysis of different spectral bands and resolution
Parameter selection of genetic algorithm(a): Iteration parameters; (b): Population size; (c): Mutation probability; (d): Crossover probability
Fig. 2. Parameter selection of genetic algorithm
(a): Iteration parameters; (b): Population size; (c): Mutation probability; (d): Crossover probability
Model prediction results of different spectral bands at the same resolution
Fig. 3. Model prediction results of different spectral bands at the same resolution
Model prediction results of different resolution at 1~2.5 μm spectral band
Fig. 4. Model prediction results of different resolution at 1~2.5 μm spectral band
Training algorithmTraining
function
RMSEPTime
/s
Bayesian regularization algorithmtrainbr1.62130
Quasi-Newton algorithmtrainbfg1.6494
Levenberg-Marquardttrainlm3.935
One-step secant algorithmtrainoss9.782
Quantitative Conjugate Gradienttrainscg43.132
Table 1. Test results of different training functions
Hidden layer
function
Output layer
function
RRMSEPtraining
time/s
logsiglogsig09.99116
purelinlogsig09.99124
tansiglogsig010168
logsigpurelin0.683.2170
tansigpurelin0.693.16171
logsigtansig0.782.61167
tansigtansig0.782.61170
purelinpurelin0.921.63130
purelintansig0.921.55130
Table 2. Test results of different node transfer function