The MKTC method clusters a data set by utilizing weakly supervised information provided on a multi-instance subset of the data set.
The MKTC method can be described in terms of two tasks:
-
A Source Task: Where a multiple kernel metric is learnt while clustering the weakly supervised multi-instance subset.
-
A Target Task: Where the learnt multiple kernel metric is used to cluster the original data set.
MKTC has a computation complexity linear in the size of the dataset, making them suitable for the clustering of large datasets.
demo.ipynb contains an example of using MKTC on the sklearn digits data set.
mktc.py contains the implemetation of MKTC.