Uncertainty quantification for deep neural networks: An empirical comparison and usage guidelines

Abstract

SummaryDeep neural networks (DNN) are increasingly used as components of larger software systems that need to process complex data, such as images, written texts, audio/video signals. DNN predictions cannot be assumed to be always correct for several reasons, amongst which the huge input space that is dealt with, the ambiguity of some inputs data, as well as the intrinsic properties of learning algorithms, which can provide only statistical warranties. Hence, developers have to cope with some residual error probability. An architectural pattern commonly adopted to manage failure prone components is the supervisor, an additional component that can estimate the reliability of the predictions made by untrusted (e.g., DNN) components and can activate an automated healing procedure when these are likely to fail, ensuring that the deep learning‐based system (DLS) does not cause damages, despite its main functionality being suspended.In this paper, we consider DLS that implement a supervisor by means of uncertainty estimation. After overviewing the main approaches to uncertainty estimation and discussing their pros and cons, we motivate the need for a specific empirical assessment method that can deal with the experimental setting in which supervisors are used, where accuracy of the DNN matters only as long as the supervisor lets the DLS continue to operate. Then we present a large empirical study conducted to compare the alternative approaches to uncertainty estimation. We distilled a set of guidelines for developers that are useful to incorporate a supervisor based on uncertainty monitoring into a DLS.

Applications of Certainty Scoring for Machine Learning Classification in Multi-modal Contexts

“…Uncertainty estimation forms a significant component of software testing for software containing deep learning elements. Previous research has found various techniques from uncertainty estimation effective in reducing defects and vulnerabilities of software containing deep neural networks [52], [65]. The main methodology for employing uncertainty quantification techniques for reducing defects in software containing deep neural networks is to use a supervisor module in a deep neural network software system that evaluates whether a given prediction by the deep neural network should be trusted or not.…”

“…The main methodology for employing uncertainty quantification techniques for reducing defects in software containing deep neural networks is to use a supervisor module in a deep neural network software system that evaluates whether a given prediction by the deep neural network should be trusted or not. Recent empirical work has found measures developed for uncertainty quantification to work well for the supervisor module [65]. As such, these uncertainty estimation techniques have been thoroughly empirically validated for their use in software engineering [65], and a Python package is even available for developers to implement many of these uncertainty quantification techniques as part of their software development [53].…”

“…Recent empirical work has found measures developed for uncertainty quantification to work well for the supervisor module [65]. As such, these uncertainty estimation techniques have been thoroughly empirically validated for their use in software engineering [65], and a Python package is even available for developers to implement many of these uncertainty quantification techniques as part of their software development [53].…”