TABLE II.

The best candidates and the corresponding regression loss (L) for u_t and v_t with respect to the number of diffusion map coordinates.

	Optimal intrinsic coordinates		Regression loss (L)
	ut	vt	ut	vt
1d	(${\varphi }_5^u$)	(${\varphi }_2^v$)	4.60 × 10−04	7.69 × 10−06
2d	(${\varphi }_1^u,{\varphi }_5^u$)	(${\varphi }_1^v,{\varphi }_2^v$)	1.40 × 10−06	1.50 × 10−06
3d	(${\varphi }_1^u,{\varphi }_4^u,{\varphi }_5^u$)	(${\varphi }_1^v,{\varphi }_2^v,{\varphi }_3^v$)	2.18 × 10−08	4.74 × 10−08
4d	(${\varphi }_1^u,{\varphi }_3^u,{\varphi }_4^u,{\varphi }_5^u$)	(${\varphi }_1^v,{\varphi }_2^v,{\varphi }_3^v,{\varphi }_4^v$)	1.64 × 10−08	5.71 × 10−09