ANN - Approximate Nearest Neighbors

Implementation of an ANN index with locality-sensitive hashing in C++, for efficient vector lookup of nearest neighbors.

Vector Projections (2D)

• LSH
• HyperCube

LSH:

Hash Vectors with amplified hash functions and use Hash Table for storage.
Get approximate nearest neighbor - Or R-nearest neighbors, with M1 metric.

HyperCube:

Hash Vectors with amplified hash functions and use Hyper Cube for storage. Get approximate nearest neighbor Or R-nearest neighbors, with M1 metric

Curve Projections (3D)

• Grid LSH
• Projections

Grid LSH:

Curve Vectorization with Grid, then calling LSH for vectors on the vectorized curves. (Poor method)

Projections:

Find all relevant traversals, and project the nearest traversals, and use LSH on vectorized traversals, instead of vectorizing Curves. (Better method for curves)

General Use functions for I\O and searching:

Helper Functions

Usage:

• Compilation
  make all
• Instructions for usage
  ./build/lsh.x
  ./build/hypercube.x

Collaborators

Konstantinos Athinaios
Theofanis Aslanidis