caokai1073/pamona

The software of Pamona, a partial manifold alignment algorithm.

pamona's Issues

Semi-supervised correspondence

Hi,

I left a similar comment for UnionCom, would it be correct to use negative values in the M discrepancy matrix to incorporate the one-to-one prior correspondece between some observations?. In such a way that Pamona would try to assign higher values in the same entry of Tij

Thanks

The visualization fails with datatype more than 12.

Hi,

In visualization.py (line 13), the styles is defined as:
styles = ['g', 'r', 'b', 'y', 'k', 'm', 'c', 'greenyellow', 'lightcoral', 'teal']

So we try to run:
Pa.Visualize(data, integrated_data, datatype, mode='UMAP') # with datatype
I got an error:

---> 11             plt.scatter(embedding[i][index,0], embedding[i][index,1], c=styles[j], s=5.)
     12         plt.title(dataset_xyz[i])
     13         if mode=='PCA':
IndexError: list index out of range 

n_shared estimation

Thanks a lot for the tool and didactic notebooks ! I did not find an example showing how to estimate the number of shared cells, would you have one to provide ?

UMAP random seed

I notice each time the visualize function might generate a different UMAP plot due to the random state being not fixed. Could you add the default random seed similar to the main Pamona optimization algorithm?