Research related to Data Engineering for Deep Learning
Bo Zhao, Konda Reddy Mopuri, Hakan Bilen ICLR, 2021 Oral (<2% of the submissions (Avg. Rating 8.33, rated top-2 in the whole conference) PDF / Codes This paper proposes a training set synthesis technique for data-efficient learning, called Dataset Condensation, that learns to condense a large dataset into a small set of informative samples for training deep neural networks from scratch. We formulate this goal as a gradient matching problem between the gradients of a deep neural network trained on the original data and our synthetic data. |