Figure 4 | Scientific Reports

Figure 4

From: Minimizing artifact-induced false-alarms for seizure detection in wearable EEG devices with gradient-boosted tree classifiers

Figure 4

Comparative analysis of artifact detection, in both sub-figures the shaded areas (brown and orange) represent the L1 and L2 memory capacity of GAP9. (a) Compares artifact detection accuracy of classifiers against model size, illustrating superior resistance and performance of a Gradient Boosting Classifier versus an Extra Trees Classifier optimized by TPOT. (b) Extends to Gradient Boosted Trees’ accuracy for three labeling methods (BC, MC, MMC) when implemented on the GAP9 processor. Star symbols denote TPOT-optimized models’ F1 scores (0.84, 0.60, 0.87 for BC, MC, MMC, respectively).

Back to article page