Lilian Weng 11/30/2018

Meta-Learning: Learning to Learn Fast

Read Original

This article explores meta-learning, a subfield of machine learning focused on creating models that can learn new concepts or skills rapidly from a few examples, similar to human learning. It defines the problem, discusses its goals in supervised learning contexts like few-shot classification, and explains the training process where models are optimized over a distribution of tasks to perform well on unseen ones.

Meta-Learning: Learning to Learn Fast

Comments

No comments yet

Be the first to share your thoughts!

Browser Extension

Get instant access to AllDevBlogs from your browser