Paper featured on GSK.ai blog!
Published:
Our work with Emma Slade at GSK.ai has been featured on the GSK.ai blog! This work considers the concept of dataset distillation, reducing the amount of data needed to train neural networks. We provably demonstrate convergence when using diffusion-based generative models to create the distilled data, and show state-of-the-art performance on very large image datasets. arXiv