Hi,
Thanks for developing the amazing tool. I was wondering if the n_clusers arg is mandatory or if users can leave it blank?
I read the tutorials and search_res() function source code, seems like the target_num, aka, the n_clusters arg is a required parameter for users to put in. However, I read in the paper that says
"To initialize
cluster centroids, we use Louvain’s method on the aggregated output matrix from
the graph convolutional layer. If the number of domains in the tissue is known, the
resolution parameter in Louvain will be set to generate the same number of spatial
domains. Otherwise, we vary the resolution parameter from 0.2 to 1.0 and select
the resolution that gives the highest Silhouette score"
Does it mean if we do not give the n_clusters number, the algo will select a good resolution on its own?
Thanks for any education!
Hi,
Thanks for developing the amazing tool. I was wondering if the
n_clusersarg is mandatory or if users can leave it blank?I read the tutorials and
search_res()function source code, seems like thetarget_num, aka, then_clustersarg is a required parameter for users to put in. However, I read in the paper that says"To initialize
cluster centroids, we use Louvain’s method on the aggregated output matrix from
the graph convolutional layer. If the number of domains in the tissue is known, the
resolution parameter in Louvain will be set to generate the same number of spatial
domains. Otherwise, we vary the resolution parameter from 0.2 to 1.0 and select
the resolution that gives the highest Silhouette score"
Does it mean if we do not give the n_clusters number, the algo will select a good resolution on its own?
Thanks for any education!