Learning to rank biological motion trajectories

Many feature transforms have been proposed for the problem of trajectory matching. These methods, which are often based on shape matching, tend to perform poorly for biological trajectories, such as cell motion, because similar biological behavior often results in dissimilar trajectory shape. Additionally, the criteria used for similarity may differ depending on the user's particular interest or the specific query behavior. We present a rank-based distance metric learning method that combines user input and a new set of biologically-motivated features for biological trajectory matching. We show that, with a small amount of user effort, this method outperforms existing trajectory methods. On an information retrieval task using real world data, our method outperforms recent, related methods by ~9%. To learn a distance metric for trajectory matching, a user selects a query trajectory (left) and provides feedback by determining which from a pair of trajectories (middle) is more similar to the query. Our algorithm uses this feedback and learns a set of feature weights to be used for trajectory matching (right). [Display omitted] ► Biologically-motivated features for cell trajectory matching. ► Leverages user input for semi-supervised distance metric learning. ► Outperforms shape-based approaches on real-world cell trajectory retrieval problem.