CellScope.cm.best_map

CellScope.cm. best_map (L1 : np.ndarray, L2 : np.ndarray)

best_map is a utility function that leverages the Hungarian algorithm [Mu57] to find the optimal mapping between two sets of numerical labels. It adjusts the predicted labels (L2) to best align with the true labels (L1), thus minimizing the mismatch between the two sets.

Parameters

• L1 (np.ndarray):

  A numerical label vector with a length of num_cell, representing the true labels for each cell.

• L2 (np.ndarray):

  A numerical label vector with a length of num_cell, representing the predicted labels for each cell.

Return

• new_L2 (ndarray):

  A numerical label vector of size (num_cell), obtained by permuting L2 using the Hungarian algorithm, making it align with L1 as closely as possible.

Notes

The best_map function utilizes the Hungarian algorithm to perform label alignment, which is commonly used in clustering tasks to measure the similarity between two label assignments. The method ensures that the predicted labels (L2) are reordered to match the true labels (L1), making it a valuable step in cluster evaluation.