Computer Vision News - February 2022

7 Agree to Disagree on building the next releases of automatic speech recognition (ASR) models for Alexa . Matthew, who is from the UK, says his work is focused on the explainability of machine learning models. “How explainability can be used to make better models, how these models can be more robust, and how this can improve the application of machine learning to healthcare,” he reveals. He is currently working on another paper applying this work specifically to healthcare data, which the team are hoping to submit to MICCAI later this year. They are also working on a solution for producing more consistent explanations to take them a step further towards finding more reliable explainable methods for deep learning in real-world applications . And we hope to see them all in the real world at MICCAI in Singapore this September! important. When you’re making financial decisions, you must be able to explain to your customers why you’ve made them. I’ve been wanting to solve these problems ever since!” Outside of this paper, Noura works on a number of sensitive applications, including healthcare more generally, as well as being the Head of Machine Learning at Evergreen , a UK healthcare organization. She also works with a local FinTech company on financial data and has a project with her PhD students looking at detecting and reducing bias in facial recognition systems. “All three are very sensitive applications, which is why we’re doing more research on explainability and exposing their vulnerabilities to try to make them more robust,” she tells us. Noura and Bashar hail from Damascus in Syria originally. Bashar’s day job is as a speech scientist at Amazon . He is working B E S T A W A R D PAPER