uniport.metrics.batch_entropy_mixing_score

uniport.metrics.batch_entropy_mixing_score(data, batches, n_neighbors=100, n_pools=100, n_samples_per_pool=100)

Calculate batch entropy mixing score

Algorithm

• 1. Calculate the regional mixing entropies at the location of 100 randomly chosen cells from all batches

• 2. Define 100 nearest neighbors for each randomly chosen cell

• 3. Calculate the mean mixing entropy as the mean of the regional entropies

• 4. Repeat above procedure for 100 iterations with different randomly chosen cells.

param data:

np.array of shape nsamples x nfeatures.

param batches:

batch labels of nsamples.

param n_neighbors:

The number of nearest neighbors for each randomly chosen cell. By default, n_neighbors=100.

param n_samples_per_pool:

The number of randomly chosen cells from all batches per iteration. By default, n_samples_per_pool=100.

param n_pools:

The number of iterations with different randomly chosen cells. By default, n_pools=100.

rtype:

Batch entropy mixing score