Navigable Graphs Python

Results

Syntetic data results:

m = 50, m0 = 50, ef_construction = 30, n = 10000, dim = 128, k = 5
ef_values = [5, 10, 20, 30, 40, 50]

Results on real data (run the code as in the example from Usage):

m = 50, m0 = 50

Dataset	Method	Average recall	Avg calc
sift10k	baseline	0.9739	246.75
sift10k	custom	0.9780	244.53
sift1m	baseline	0.8074	420.43
sift1m	custom	0.7943	413.52

Usage

Python based research tool for studying navigable graphs for nearest neighbour search

Using the SIFT dataset:

python navigable-graphs.py --dataset sift

Using synthetic data with 3D vectors:

python navigable-graphs.py --dataset synthetic --K 20 --k 5 --dim 3 --n 500 --nq 100 --ef 20 --M 2