Scientific Reports

Table 4 Low vs high valence classification accuracies obtained with EEGNet.

From: Two-dimensional CNN-based distinction of human emotions from EEG channels selected by multi-objective evolutionary algorithm

S	Channels
S	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
1	0.812	0.929	0.933	0.938	0.946	0.975		0.992		1.000
2	0.592	0.721	0.792	0.833	0.871	0.875	0.879	0.908	0.917
3	0.733	0.833	0.967	0.975	1.000
4		0.792	0.825	0.838	0.875	0.9	0.908		0.921			0.938		0.942
5	0.558	0.649	0.814	0.86	0.874	0.881	0.909	0.94
6	0.698		0.827	0.867	0.882	0.933		0.941			0.976		0.980		0.984
7	0.564	0.76	0.796	0.911			0.964					0.978	0.991		1.000
8	0.650	0.675	0.717	0.783	0.838		0.862	0.912	0.929	0.938
9	0.693	0.778	0.818	0.84		0.849	0.862		0.876	0.884	0.920
10	0.589	0.674	0.785	0.800	0.852			0.878		0.915		0.933		0.944
11	0.644	0.693	0.804	0.844	0.849	0.871		0.911
12	0.905	0.962	0.981	1.000
13	0.922	1.000
14	0.615	0.738	0.774		0.826	0.831		0.856	0.872	0.908		0.928
15	0.775	0.867	0.898		0.916		0.94	0.979
16	0.800	0.890	0.900	0.910	0.930		0.943		0.96	0.977		0.983	0.997
17	0.716		0.738	0.751	0.756		0.778			0.782	0.800
18		0.791	0.844	0.849	0.88		0.893		0.907			0.924			0.929
19	0.554	0.744	0.867		0.877	0.913	0.938	0.944	0.954
20	0.822	0.904	0.941	0.956	0.978	1.000
21	0.633	0.825	0.933	0.958	0.975		0.983		1.000
22	0.684	0.809	0.831	0.858	0.871	0.876	0.889	0.907	0.916
23	0.641	0.754	0.81	0.846	0.892			0.923			0.944	0.969		0.974	0.979
24	0.648	0.876	0.914	0.971		0.981	1.000
25	0.747	0.793	0.847	0.940	0.993	1.000
26	0.569	0.725	0.839	0.863	0.914	0.929	0.957	0.984	0.996
27	0.692	0.856	0.903		0.918	0.949	0.985	0.990		0.995
28	0.660	0.737	0.787			0.803	0.807	0.823	0.903	0.907					0.930
29	0.621	0.713	0.774	0.790	0.841		0.877	0.985					0.995
30	0.627	0.740		0.777	0.84		0.860	0.883	0.940
31	0.713	0.760	0.843		0.867	0.877	0.917	0.947		0.960			0.977
32	0.772	0.856	0.928	0.967	0.994	1.000

Accuracies obtained in the Pareto-front for the first 1-15 channels selected by NSGA-II.
The subjects with the highest accuracies per channel set are indicated in bold.

Back to article page

Search

Advanced search

Quick links