Automatic Tuning of Interactive Perception Applications

TL;DR

This paper presents an online automatic tuning method for interactive perception applications that learns application models to optimize fidelity within latency constraints, reducing manual tuning effort.

Contribution

It introduces a novel online learning approach that models application performance and efficiently finds optimal parameters for interactive perception systems.

Findings

Latency models learned accurately online

Application structure reduces learning complexity

Achieves 90% of optimal fidelity with minimal parameter exploration

Abstract

Interactive applications incorporating high-data rate sensing and computer vision are becoming possible due to novel runtime systems and the use of parallel computation resources. To allow interactive use, such applications require careful tuning of multiple application parameters to meet required fidelity and latency bounds. This is a nontrivial task, often requiring expert knowledge, which becomes intractable as resources and application load characteristics change. This paper describes a method for automatic performance tuning that learns application characteristics and effects of tunable parameters online, and constructs models that are used to maximize fidelity for a given latency constraint. The paper shows that accurate latency models can be learned online, knowledge of application structure can be used to reduce the complexity of the learning task, and operating points can be found that achieve 90% of the optimal fidelity by exploring the parameter space only 3% of the time.