Projects per year
Abstract / Description of output
As the state-of-the-art machine learning methods in many fields rely on larger datasets, storing datasets and training models on them become significantly more expensive. This paper proposes a training set synthesis technique for data-efficient learning, called Dataset Condensation, that learns to condense large dataset into a small set of informative synthetic samples for training deep neural networks from scratch. We formulate this goal as a gradient matching problem between the gradients of deep neural network weights that are trained on the original and our synthetic data. We rigorously evaluate its performance in several computer vision benchmarks and demonstrate that it significantly outperforms the state-of-the-art methods. Finally we explore the use of our method in continual learning and neural architecture search and report promising gains when limited memory and computations are available.
Original language | English |
---|---|
Title of host publication | International Conference on Learning Representations (ICLR 2021) |
Number of pages | 20 |
Publication status | E-pub ahead of print - 29 Mar 2021 |
Event | Ninth International Conference on Learning Representations 2021 - Virtual Conference Duration: 4 May 2021 → 7 May 2021 https://iclr.cc/Conferences/2021/Dates |
Conference
Conference | Ninth International Conference on Learning Representations 2021 |
---|---|
Abbreviated title | ICLR 2021 |
City | Virtual Conference |
Period | 4/05/21 → 7/05/21 |
Internet address |
Fingerprint
Dive into the research topics of 'Dataset Condensation with Gradient Matching'. Together they form a unique fingerprint.Projects
- 1 Active
-
Visual AI: An Open World Interpretable Visual Transformer
Engineering and Physical Sciences Research Council
1/12/20 → 30/11/26
Project: Research