AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Using Rules to Generate More Data for Deep Learning Systems
The way you describe that resignates really strongly with the conversation i had just the other with peter beale. He made this really interesting point that if you say, oca, we want to incorporate this knowledge that we have about a problem, he one way to do it is to build rules into your system. But what is said that struck me as really interesting was another way to do that is to use your rules to generate more data for your deep learning systems. And so in that way, you're training them on the cases that you know a lot about, but still not having to take on the technical debt of having a lot of rules. Maybe that is the kind of glue