Going beyond Imitation Learning: Learning from Non-Traditional Sources of Human Data

Prof. Dorsa Sadigh presents a set of techniques to address some of the challenges of learning from non-traditional sources of data, i.e., suboptimal demonstrations, rankings, and play data. Sadigh will first introduce the basics of imitation learning then will talk about her lab's confidence-aware imitation learning approach that simultaneously estimates a confidence measure over demonstrations and the policy parameters.