Be certain about model uncertainty with our guide to probabilistic predictions
At HAL24K we know the value of communicating our models’ uncertainty to our customers, and a few months ago we published an article detailing how we go about generating this uncertainty for deep learning models we have already developed. However, what if – as some of the responses to that article asked – we instead set out to develop a new deep learning model that can predict its own uncertainty?
In the latest post in our TechBlog, Sam Blake, one of our senior data scientists, guides you through the process of developing these kinds of models, from choosing an initial uncertainty distribution through to visualizing the predictions of the trained model.
You can find the article here.