Sustain the Smartness: from Smart Things to Sustainable Smart Things.

Retraining is an essential process for the sustainability of any production Machine Learning (ML) model and/or any software with ML based components, as retraining addresses the problem of data shift. However, uncertainty from different sources makes the successful retraining a challenging task. Here, we provide an outline of a multi-armed bandit based decision framework, which can address the uncertainty related to a retraining framework of any production ML model.