Condensing Graphs via One-Step Gradient Matching. (arXiv:2206.07746v1 [cs.LG])

Web: http://arxiv.org/abs/2206.07746

As training deep learning models on large dataset takes a lot of time and
resources, it is desired to construct a small synthetic dataset with which we
can train deep learning models sufficiently. There are recent works that have
explored solutions on condensing image datasets through complex bi-level
optimization. For instance, dataset condensation (DC) matches network gradients
w.r.t. large-real data and small-synthetic data, where the network weights are
optimized for multiple steps at each outer iteration. However, existing
approaches have …

arxiv gradient graphs lg