Based on the comparability, the data from electronic tongue
and nose were reduced to three dimensions by LLE algorithm.
Once the value of d=3 by LLE algorithm was identified, the
optimal value of k=7 by LLE algorithm was found, which
corresponded to the smallest values of error shown in Fig. 5a.
It was seen from Fig. 4b, d that two types of rice cannot be
demarcated completely by LLE, but samples of the same type
have a more intensive distribution in dimension compared
with PCA.