TABLE III.

Supervised classification accuracy for parameter recovery using a trained linear SVM with various input feature vectors. In some cases, the dimensionality of feature vectors has been reduced using PCA.

Summary	Feature	Dimension	Accuracy (%)
Order parameters	P(t)	87	57.7
	Mang(t)	87	34.4
	Mabs(t)	87	68.0
	DNN(t)	87	91.1
	All	4 × 87	89.2
	All (PCA)	87	69.6
	P(t) (PCA)	3	46.7
	Mang(t) (PCA)	3	30.0
	Mabs(t) (PCA)	3	58.8
	DNN(t) (PCA)	3	81.5
	All (PCA)	3	68.6
TDA (position)	b0	200 × 87	97.0
	b1	200 × 87	93.7
	b0 and b1	2 × 200 × 87	96.4
	b0 (PCA)	87	96.2
	b1 (PCA)	87	95.2
	b0 & b1 (PCA)	87	96.2
	b0 (PCA)	3	93.0
	b1 (PCA)	3	79.4
TDA (time-delayed position)	b0 & b1 (PCA)	3	93.1
	b0	200 × 86	99.6
	b1	200 × 86	99.3
	b0 & b1	2 × 200 × 86	99.1
	b0 (PCA)	87	99.7
	b1 (PCA)	87	99.9
	b0 & b1 (PCA)	87	99.7
	b0 (PCA)	3	89.7
	b1 (PCA)	3	82.8
	b0 & b1 (PCA)	3	89.6