Cluster with custom metric

Hi

I want to cluster my data using a custom distance metric, so I have some data vector V and a matrix with distances between the rows of V, pairwise. How to do it? I saw that linkage might solve it but I have no idea how to pass my metric there

Answer 1

[arushi]

Hi Teresa,

To cluster your data using a custom distance metric in MATLAB, you can use the linkage function, which supports clustering based on a precomputed distance matrix. If you already have a matrix of pairwise distances, you can pass it directly to the linkage function.

Steps to Cluster Using a Custom Distance Matrix

1. Compute the Distance Matrix: If you haven't already, compute the pairwise distance matrix for your data vector V. This matrix should be a square, symmetric matrix where the element at position (i, j) represents the distance between the ith and jth elements of V.
2. Convert the Distance Matrix to a Vector: The linkage function requires the distance matrix to be in vector form, as returned by the pdist function.
3. Perform Hierarchical Clustering: Use the linkage function with your custom distance vector. You can specify the linkage method (e.g., 'single', 'complete', 'average').
4. Form Clusters: Use the cluster function to form clusters from the hierarchical tree.

Hope this helps.