Dataset Distillation by Matching Training Trajectories

Dataset distillation is the task of synthesizing a small dataset such that a model trained on the synthetic set will match the test accuracy of the model trained on the full dataset. In this paper, we propose a new formulation that optimizes our distilled data to guide networks to a similar state as those trained on real data across many training steps. Given a network, we train it for several iterations on our distilled data and optimize the distilled data with respect to the distance between the synthetically trained parameters and the parameters trained on real data. To efficiently obtain the initial and target network parameters for large-scale datasets, we pre-compute and store training trajectories of expert networks trained on the real dataset. Our method handily outperforms existing methods and also allows us to distill higher-resolution visual data.

PDF Abstract CVPR 2022 PDF CVPR 2022 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Dataset Distillation - 1IPC CIFAR-10 MTT Classification Accuracy 46.3±0.8 # 7
Dataset Distillation - 1IPC CIFAR-100 MTT Classification Accuracy 24.3±0.3 # 7
Dataset Distillation - 1IPC CUB-200-2011 MTT Classification Accuracy 2.2±0.1 # 3
Dataset Distillation - 1IPC TinyImageNet MTT Classification Accuracy 8.8±0.3 # 6


No methods listed for this paper. Add relevant methods here