Semi-supervised correspondence

[Andresd45]

Hi,

Is it possible to run the code including some one-to-one correspondece?, it is mention in the paper but I can no see how to do it with the code.

Thanks!

[caokai1073]

Hi,

Thanks for your interest. Semi-supervised mode is not involved in our code but easy to implement.
For example, if cell i in dataset X and cell j in dataset Y have correspondence, we can initialize matching matrix $F$ in Prime_Dual function with corresponding $F_{ij}$ as 1 and others still as 0 (line 248 in UnionCom.py). Besides, after updating F in line 295, $F_{ij}$ also need to be modified as 1 instead of new updated value, which means we need to add F[i][j]=1 to line 296 in UnionCom.py.

Best,
Kai