Journal, 6 November 2018
Zero-shot learning: Research team at Facebook is currently developing a model that uses text to learn to identify images it has not seen before. This technology, if fully developed, might border on creepy. But, I’m imagining how this could help handle imbalanced data, say by describing how cancer would look like on MRI scan. So many possibilities, no?
P.S.: The post came out late because I was out late to see Moore Moss Rutter at the Slaughtered Lamb. Check them out if you haven’t heard of them!