Input responsiveness: using canary inputs to dynamically steer approximation

Abstract

This paper introduces Input Responsive Approximation (IRA), an approach that uses a canary input-a small program input carefully constructed to capture the intrinsic properties of the original input-to automatically control how program approximation is applied on an input-by-input basis. Motivating this approach is the observation that many of the prior techniques focusing on choosing how to approximate arrive at conservative decisions by discounting substantial differences between inputs when applying approximation. The main challenges in overcoming this limitation lie in making the choice of how to approximate both effectively (e.g., the fastest approximation that meets a particular accuracy target) and rapidly for every input. With IRA, each time the approximate program is run, a canary input is constructed and used dynamically to quickly test a spectrum of approximation alternatives. Based on these runtime tests, the approximation that best fits the desired accuracy constraints is selected and applied to the full input to produce an approximate result. We use IRA to select and parameterize mixes of four approximation techniques from the literature for a range of 13 image processing, machine learning, and data mining applications. Our results demonstrate that IRA significantly outperforms prior approaches, delivering an average of 10.2× speedup over exact execution while minimizing accuracy losses in program outputs.

An Efficient Monte Carlo-Based Probabilistic Time-Dependent Routing Calculation Targeting a Server-Side Car Navigation System

“…Second, in the iterative phase, it suggests a reactive approach rather than a proactive one, that we already discussed to be necessary for the specific PTDR problem. Another two step approach is the one presented in [24]. In this case, the authors suggest building from each input a small canary -a statistically representative subset of the actual input -that is used to perform a parameter exploration at runtime.…”

One Size Does Not Fit All: Quantifying and Exposing the Accuracy-Latency Trade-Off in Machine Learning Cloud Service APIs via Tolerance Tiers

“…Approximate Computing Approximation techniques can be static [51] or dynamic [52], [53]. Our work is the latter, treating training data as canary inputs [54]. At its core Tolerance Tiers is a domain specific approximate computing technique.…”

Towards Approximate Mobile Computing

“…In most situations, we can evaluate the quality only if the result of a perfectly accurate computation is available, defying the purpose of approximation. Laurenzao et al show that in image approximation it suffices to evaluate the result quality on small representative snippets of data [14], yet, this might not generalize to other domains. In addition, the app needs to have the information about the current context in order to adapt to it.…”