RoarGraph: A Projected Bipartite Graph for Efficient Cross-Modal Approximate Nearest Neighbor Search | MysteryANN
The codes for the NIPS 2023 challenge are available in separate branches. The main branch is the codebase of the RoarGraph paper. It also 🏆 Winning NeurIPS' Competition Track: Big ANN, Practical Vector Search Challenge. (OOD Track) (Our other solution won the Sparse Track).
You can find the codes for the NIPS 2023 competition in other branches, NOT the main branch, even if they have similar designs from the same research project. The paper version provides a better index.
I also prepared a tidy code repository for the VLDB paper. Please refer to https://github.com/matchyc/RoarGraph for detailed instructions, and avoid using the main branch to reproduce paper, you may face issues.
This code builds upon the NSG repo and incorporates other open-source implementations.
If you wish, please leave a message if you plan to use this idea.
For questions or inquiries, feel free to reach out to Meng Chen at [email protected]