README.md

NPG_nsw

Compile on Linux

cd NPG_nsw/n2/
make
cd examples/cpp/
make

Build NPG_nsw Index

First:

cd NPG_nsw/n2/examples/cpp/

Then:

./index data_file save_graph MaxM0 efConstruction

Meaning of the parameters:

<data_file> is the path of the original object set.
<save_graph> is the path of the NPG_nsw to be saved.
<MaxM0> is the 'K' of kNN graph.
<efConstruction> is the parameter contolling the graph quality, larger is more accurate but slower.

Search on NPG_nsw

./search graph_path query_file groundtruth_file

Meaning of the parameters:

<graph_path> is the path of the pre-built NPG_nsw.
<query_file> is the path of the query object.
<groundtruth_file> is the path of the groundtruth data.

Hybrid Queries on NPG_nsw ("first vector similarity search, then attribute filtering")

./hybrid_search graph_path data_path query_path base_att_path query_att_path groundtruth_path

Meaning of the parameters:

<graph_path> is the path of the pre-built NPG_nsw.
<data_path> is the path of the original object set.
<query_path> is the path of the query object.
<base_att_path> is the path of the corresponding structured attributes of the original objects.
<query_att_path> is the path of the corresponding structured attributes of the query object.
<groundtruth_path> is the path of the groundtruth data.