Timezone: »

 
Poster
TPU-KNN: K Nearest Neighbor Search at Peak FLOP/s
Felix Chern · Blake Hechtman · Andy Davis · Ruiqi Guo · David Majnemer · Sanjiv Kumar

Thu Dec 01 09:00 AM -- 11:00 AM (PST) @ Hall J #910
in Poster Session 5 »

This paper presents a novel nearest neighbor search algorithm achieving TPU (Google Tensor Processing Unit) peak performance, outperforming state-of-the-art GPU algorithms with similar level of recall. The design of the proposed algorithm is motivated by an accurate accelerator performance model that takes into account both the memory and instruction bottlenecks. Our algorithm comes with an analytical guarantee of recall in expectation and does not require maintaining sophisticated index data structure or tuning, making it suitable for applications with frequent updates. Our work is available in the open-source package of Jax and Tensorflow on TPU.

Keywords: TPU · accelerator · roofline model · K-nearest neighbor search · Approximate nearest neighbor search
[ Paper]  [ Poster]  [ OpenReview] 

Author Information

Felix Chern (Google Inc)
Blake Hechtman (Google)
Andy Davis
Ruiqi Guo (Google)
David Majnemer (Google)
Sanjiv Kumar (Google Research)

More from the Same Authors