Additive Quantization

This is an implementation of Additive Quantization by Julieta as described in

Artem Babenko and Victor Lempitsky: Additive Quantization for Extreme Vector Compression, CVPR 14.

TODOs

• The code for encoding under AQ_encoding.m can surely benefit from a C/C++ implementation.
• The codeboook update under AQ_update_codebooks.m runs a least-squares uptimization on sparse matrix of code assignments. My implementation currently creates a full matrix and then makes it sparse, which is quite readable but not very good for speed/memory.