Implementation Idea and pseudo-code analysis of HC-fused

[yun-lai-yun-qu]

Hi author, I read the pseudocode of the paper “A hierarchical clustering and data fusion approach for disease subtype discovery” you mentioned, but didn't quite understand how clusters c1 and c2 are obtained, and how in each view, each data is reused until there is only one cluster per view?

[pievos101]

Hi! Thanks for your post.

In the paper you mentioned the clusters per view are inferred by the silhouette coefficient. These clusters are then represented as a binary matrix which is the input for the introduced fusion algorithm. This generates a similarity matrix which to this end is just another view, but with pairwise distances already computed.

We hope this helps.

All the best

[yun-lai-yun-qu]

@pievos101 Yes, I know what you mean. Now I am confused about how to get the similarity matrix according to different binary matrices, that is, the pseudo-code Algorithm 1 in the paper. I do not understand this process, especially how to get c1 and c2 in Algorithm 1

[pievos101]

Hi!

The idea is simple. A dendrogram is build using a bottom-up approach. The fusion algorithm can choose the distances from ANY binary matrix in order to decide which samples or clusters to fuse. During that process it is counted how many times two samples appear in the same cluster. Note, the AND binary matrix is preferred ...

All the best, Bastian