vector quantization?

[Hristo Stoyanov]

Hi all,

I am searching for a Java library/code that can compress float[] or double[] into 8-bit (byte[]) or even 4-bit encoded vector, with some loss of precision, but also provide API for basic vector operations over the compressed vectors. Can Roaring Bitmaps do something like this?

[Daniel Lemire]

You can probably use RoaringBitmap instances as part of the solution, but we don't offer anything that looks like what you are describing.