Robust L-Isomap with a Novel Landmark Selection Method

Figure 6

Comparison of RL-Isomap with Robust Kernel Isomap for the case of Swiss roll data. (a) Neighbourhood graph by Robust Kernel Isomap. (b) Neighbourhood graph after removing the edges whose edge density values are below the threshold by RL-Isomap. (c) The distribution of the edge density values of edges in the new neighbourhood graph. (d) The landmark candidate points and the landmark points. (e) Embedding result by Robust Kernel Isomap. (f) Embedding result by RL-Isomap.

(a)

(b)

(c)

(d)

(e)

(f)