How do you train a machine learning model on something that's not there? 🫥 This research note shows how to use low-capacity networks to, for example, predict the severity of an infection on a scale based on information on just rough outcomes in previous cases. www.epcsirmaz.co.uk/tech_blog/20...