Active learning is a branch in machine learning that uses a judicial choice of data to label in order to train an ML system. Such an approach is needed when labeling is difficult or expensive and it is impossible to label a large data set.
In this talk we show that the active learning problem has strong links to the problem of sensor placement and that this problem can optimally be solved by A optimal experimental design. We call the approach A optimal active learning.
Back to Workshop II: PDE and Inverse Problem Methods in Machine Learning