skip to primary navigationskip to content
 

High-Quality Prediction Intervals for Deep Learning: A Distribution-Free, Ensembled Approach

last modified Feb 21, 2018 10:36 AM
February Paper on High-Quality Prediction Intervals for Deep Learning: A Distribution-Free, Ensembled Approach by Tim Pearce, Mohamed Zaki, Alexandra Brintrup, Andy Neely

[Paper]

Deep neural networks (NNs) have caused great excitement due to the step-changes in performance they have delivered in a variety of applications. However, their appeal in industry can be inhibited by an inability to quantify the uncertainty of their predictions. To take a prognostics example, a typical NN might predict that a machine will fail in 60 days. It is unclear from this point prediction whether the machine should be repaired immediately, or whether it can be run for another 59 days. However, if the NN could output a prediction interval (PI) of 45-65 days with 99% probability, timing of a repair could easily be scheduled. In this paper, we develop a method for doing exactly this - the quantification of uncertainty in deep learning using PIs. We derive a method based on the assumption that high-quality PIs should be as narrow as possible, whilst still capturing a given proportion of data. The method is general, applicable to any data-driven task where a continuous value needs to be predicted, and it is important to know the uncertainty of that prediction. Examples include the forecasting of precipitation, energy load, financial metrics, and traffic volume. The method is tested on ten real-world, open-source datasets. The proposed method is shown to outperform current state-of-the-art uncertainty quantification methods, reducing average PI width by around 10%.

RSS Feed Latest news

Industry Day 2020 | What we learnt

Dec 06, 2020

What is the future of services in the new pandemic-driven era? This was the topic explored at our 2020 Industry Day by senior executives from L'Oréal, Microsoft Research, Manchester United and start-up, Fairjungle.

Understanding business models in the construction sector

Nov 25, 2020

What’s standing in the way of offsite manufacturing? For decades, the construction sector has been hailing it as the next big thing but we have yet to see it really taking off. Why is that, when the technologies and processes already exist? Dr Zakaria Dakhli believes it is due to a fundamental incompatibility between business models and it is only when this has been fully understood that the long-awaited transformation can take place.

A machine learning approach to quality control

Nov 14, 2020

Most of the products we take for granted contain huge numbers of components assembled in multiple stages by different manufacturers. Quality control is vital throughout the assembly process with rigorous testing required at every step.

Digital twins: driving business model innovation

Nov 06, 2020

For B2B firms struggling to reap the rewards of digitalisation, could digital twins be the way forward? In this article, CSA's Dr Erika Pärn and colleagues from the Fraunhofer Institute and Technische Universität in Dortmund, explore the relationship between this emerging technology and business model innovation.

Making business model innovation happen: the Business Model Cohesiveness Scorecard

Oct 30, 2020

Digital transformation is failing to live up to its hype, at least as far as productivity gains are concerned. If, as research suggests, a lack of business model innovation is the main culprit, we need a way of making it happen. Dr Chander Velu thinks a balanced scorecard approach could be the answer.

View all news