DeepEverest: Accelerating Declarative Top-K Queries for Deep Neural Network Interpretation

TL;DR

DeepEverest is a system that significantly speeds up interpretation by example queries over deep neural network activations, using efficient indexing and algorithms proven to be instance optimal, outperforming existing methods.

Contribution

It introduces a novel indexing technique and an optimized query execution algorithm for DNN interpretation, with proven instance optimality and substantial performance improvements.

Findings

Up to 63x faster query execution

Uses less than 20% of storage compared to full materialization

Outperforms other methods on multi-query workloads

Abstract

We design, implement, and evaluate DeepEverest, a system for the efficient execution of interpretation by example queries over the activation values of a deep neural network. DeepEverest consists of an efficient indexing technique and a query execution algorithm with various optimizations. We prove that the proposed query execution algorithm is instance optimal. Experiments with our prototype show that DeepEverest, using less than 20% of the storage of full materialization, significantly accelerates individual queries by up to 63x and consistently outperforms other methods on multi-query workloads that simulate DNN interpretation processes.