Sequential Subset Matching for Dataset Distillation