1 min to read
Metric Based Meta Learning
Matching Networks, Prototypical Networks, Siamese Networks, Relation Networks
Video
Meta learning can be described as “learning to learn.” It means model learns the learning strategy. There is a three main approach in meta learning: metric-based, model-based, and optimization-based. Metric-based approach is easy to use and can be used in any model, so it is popular and well-studied method. In this seminar I briefly introduced metric-based meta learning methods: Matching Netwroks, Prototypical Networks, and Relation Networks.
Comments