Weighted Vectors
An undergraduate research project to explore how to speed up similarity searches using hashing algorithms.
We developed a general framework that can be used to augment discrete locality sensitive hashing schemes in order to produce more accurate similarity estimates, without sacrificing much in terms of computational time.
A preprint version of the paper can be found here. The paper was published in the 4th Chinese Conference on Pattern Recognition and Computer Vision.
Results (our framework is represented by the blue and yellow curves):
