Currently can't use MMD integration with integrate_on parameter as the sampler takes group=patient_column. If integrate_on is specified, check if there are multiple groups in the current batch and use the same number of group batches from each group for MMD calculation. If one group has more batches than the other, use the smallest number of group batches and subset others randomly.
Currently can't use MMD integration with
integrate_onparameter as the sampler takesgroup=patient_column. Ifintegrate_onis specified, check if there are multiple groups in the current batch and use the same number of group batches from each group for MMD calculation. If one group has more batches than the other, use the smallest number of group batches and subset others randomly.