Efficient Large-scale Approximate Nearest Neighbor Search on the GPU
This work addresses the problem of fast and scalable nearest neighbor search for high-dimensional data in real-world applications, representing a significant advancement rather than an incremental improvement.
The paper tackles efficient approximate nearest neighbor search in high-dimensional spaces by proposing a Product Quantization Tree (PQT) method that reduces vector comparisons and includes a parallelizable re-ranking technique, achieving superior GPU performance over CPU methods on standard datasets for applications like video loop-closing.
We present a new approach for efficient approximate nearest neighbor (ANN) search in high dimensional spaces, extending the idea of Product Quantization. We propose a two-level product and vector quantization tree that reduces the number of vector comparisons required during tree traversal. Our approach also includes a novel highly parallelizable re-ranking method for candidate vectors by efficiently reusing already computed intermediate values. Due to its small memory footprint during traversal, the method lends itself to an efficient, parallel GPU implementation. This Product Quantization Tree (PQT) approach significantly outperforms recent state of the art methods for high dimensional nearest neighbor queries on standard reference datasets. Ours is the first work that demonstrates GPU performance superior to CPU performance on high dimensional, large scale ANN problems in time-critical real-world applications, like loop-closing in videos.