Table 1 The training loss and test accuracy of different CNN classifiers trained by different 3D-EEM spectra

From: Attention improvement for data-driven analyzing fluorescence excitation-emission matrix spectra via interpretable attention mechanism

Model structure

Dataset

Training loss

Test accuracy

Raw CNN classifier

Raw data

0.092 ± 0.013

89.8 ± 2.5%

Raw CNN classifier

Cut data

0.116 ± 0.019

91.3 ± 1.2%

CNN classifier with CBAM

Cut data

0.103 ± 0.011

91.2 ± 1.2%

Back to article page