sift1M data

[jianshu93]

Hello James,

Would it be interesting to also provide a large dataset, the sift 1 million dataset, which is often used to test large scale performance of NNS algorithms.

Thanks,

Jianshu

[jlmelville]

This is an interesting idea, thanks for suggesting it. I don't see it used a lot in dimensionality reduction methods because there aren't any interesting labels, but it's at least close in spirit to those sorts of datasets.