Calvin Lab Auditorium
Clustering is typically studied in the unsupervised learning setting. But in many applications, such as personalized recommendations, one cannot reach the optimal clustering without interacting with the end user. In this talk, I will describe a recent framework for interactive clustering with human in the loop. The algorithm can interact with the human in stages and receive limited, potentially noisy feedback to improve the clustering. I will present our preliminary results in this model and mention open questions.