- 2017-04-30 09:33:47
Approximation: what are the approximation abilities of function classes commonly ued for machine learning? how do we design appropriate representations? How many functions are there in the class? Universal approximators? how to improve approximation?
Estimation: how much data is required for traning? how do we use data appropriately to learn? overfitting and generalization. What are the problem parameters? how to balance complexity to get the best performance?
Optimization: how much computation is required to do learning? how do we do the learning? how to find the best hypothesis? how to reduce error?
Supervised learning: batch learning, online learning
说明 · · · · · ·