Closed farhadmd7 closed 2 months ago
Thanks! Looks good to me. Please always ensure that the order of the samples is the same in the model and in the class. It is a very typical mistake (I made it several times already 😅). Here, I fixed it and also gave a bit more interpretable name to a different distances calculation approach
2 different distances are calculated here and stored
first stored like following only if calculate_representations is set true when calling
calculate_distance_matrix(calculate_representations= True)
, here first patient embeddings are obtained from MRVI, then aggregated and then distance matrix is calculatedself.adata.uns[self.DISTANCES_UNS_KEY + "_approach1"] = distances
second approach which is app. suggested in the MRVI paper, first distances are calculated by MRVI then aggregated which is stored by defaultl in
self.adata.uns[self.DISTANCES_UNS_KEY