Title: Learning How to Learn is Learning With Point Sets
Abstract: This paper develops a simple interpretation of learning how to learn: it is ordinary learning, but from point sets, rather than points. This is an alternative to the Bayesian viewpoint of “learning a prior” (Baxter, 1996b). The idea behind learning how to learn is to partition the data into separate learning tasks, learn a model for the tasks, and then apply this model to new tasks. Ordinary learning methods do the same thing, but with individual data points as the “tasks.” The partitioning for learning how to learn can be recovered automatically, generalizing the idea of “task clustering” (Thrun and O’Sullivan, 1996). Virtually all existing algorithms fit naturally into this unifying framework, including learning a distance metric and learning internal representations.
Publication Year: 1999
Publication Date: 1999-07-07
Language: en
Type: article
Access and Citation
Cited By Count: 24
AI Researcher Chatbot
Get quick answers to your questions about the article from our AI researcher chatbot