KNN PCA

Use as is on Deng 2014

1. We performed PCA first and then computed the distance matrix on the PC loadings.
2. We first perform a kNN smoothing on the covariance matrix (with number of nearest neighbors taken to be 3 as per Euclidean distance ) and then we apply the PCA based on the modified or the smoothed covariance matrix.

1	2

Use with MST on Deng 2014

1. We perform MST (Minimal Spanning Tree) based on the PC loadings data without kNN local smoothing
2. We perform MST (Minimal Spanning Tree) based on the PC loadings data with kNN local smoothing

1	2

Use as is on simulated data

Method	High signal	Noisy samples
PCA
kNN PCA

Method	Noisy features	Noisy samples & features
PCA
kNN PCA

Use with MST on simulated data

Method	High signal	Noisy samples
PCA
kNN PCA

Method	Noisy features	Noisy samples & features
PCA
kNN PCA