Interaction for Machine Learning

Опубликовано 8 августа 2016, 22:49

The traditional role of the human operator in machine learning problems is that of a batch labeler, whose work is done before the learning even begins. However, humans can provide guidance to a learning system via a richer set of inputs and interact directly with the learning algorithm as it learns. Active research problems in this space include designing interactions for acquiring human guidance, active and life-long learning, interactive clustering, query by selection, learning to rank, and effective crowdsourcing. Addressing these problems requires contributions from multiple disciplines such as human-computer interaction, artificial intelligence, and machine learning. In this session, we bring together researchers from different disciplines and identify how richer forms of interaction with humans could help machine learning systems, in terms of both surveying the major paradigms and sharing information about new work in this area. Through a combination of presentations and discussions, we hope to gain a better understanding of the available algorithms and best practices, of their inherent limitations, and of challenges and opportunities for future research in this space.