TPU-KNN: K Nearest Neighbor Search at Peak FLOP/s

06/28/2022
by   Felix Chern, et al.
15

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.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
01/04/2023

Automating Nearest Neighbor Search Configuration with Constrained Optimization

The approximate nearest neighbor (ANN) search problem is fundamental to ...
research
06/28/2019

PUFFINN: Parameterless and Universally Fast FInding of Nearest Neighbors

We present PUFFINN, a parameterless LSH-based index for solving the k-ne...
research
12/02/2019

GGNN: Graph-based GPU Nearest Neighbor Search

Approximate nearest neighbor (ANN) search in high dimensions is an integ...
research
04/09/2008

Fast k Nearest Neighbor Search using GPU

The recent improvements of graphics processing units (GPU) offer to the ...
research
11/05/2021

SPANN: Highly-efficient Billion-scale Approximate Nearest Neighbor Search

The in-memory algorithms for approximate nearest neighbor search (ANNS) ...
research
09/11/2015

A reliable order-statistics-based approximate nearest neighbor search algorithm

We propose a new algorithm for fast approximate nearest neighbor search ...
research
08/12/2018

Reconfigurable Inverted Index

Existing approximate nearest neighbor search systems suffer from two fun...

Please sign up or login with your details

Forgot password? Click here to reset