Synthesizing Trajectory Queries from Examples

TL;DR

This paper introduces Quivr, a framework that automatically synthesizes trajectory queries from examples, effectively handling real-world data's fuzzy nature and tuning challenges, with demonstrated efficiency and accuracy.

Contribution

The paper presents Quivr, a novel framework with new pruning and semantics techniques for efficient, accurate trajectory query synthesis from examples.

Findings

Successfully synthesizes accurate queries for 17 benchmark tasks.

Significantly reduces synthesis time through proposed optimizations.

Demonstrates effectiveness on real-world trajectory data.

Abstract

Data scientists often need to write programs to process predictions of machine learning models, such as object detections and trajectories in video data. However, writing such queries can be challenging due to the fuzzy nature of real-world data; in particular, they often include real-valued parameters that must be tuned by hand. We propose a novel framework called Quivr that synthesizes trajectory queries matching a given set of examples. To efficiently synthesize parameters, we introduce a novel technique for pruning the parameter space and a novel quantitative semantics that makes this more efficient. We evaluate Quivr on a benchmark of 17 tasks, including several from prior work, and show both that it can synthesize accurate queries for each task and that our optimizations substantially reduce synthesis time.