The MNIST and USPS datasets in the link contain far fewer samples than the original datasets, so they should be the remaining ones after selection. Could you please explain how these two datasets were selected and what the selection strategy or criteria are? Thank you!
The MNIST and USPS datasets in the link contain far fewer samples than the original datasets, so they should be the remaining ones after selection. Could you please explain how these two datasets were selected and what the selection strategy or criteria are? Thank you!