Automatic Inner-loop Optimization Boosts Dataset Distillation Performance

AT‑BPTT improves dataset distillation accuracy by 6.16%, speeds inner‑loop training 3.9× and cuts memory use by 63% on CIFAR‑10, CIFAR‑100 and ImageNet‑1K. Read more: getnews.me/automatic-inner-loop-opt... #datasetdistillation #atbptt