As the availability and importance of relational data -- such as the friendships summarized on a social networking website -- increases, it becomes increasingly important to have good models for such data. The kinds of latent structure that have been considered for use in predicting links in such networks have been relatively limited. In particular, the machine learning community has focused on latent class models, adapting nonparametric Bayesian methods to jointly infer how many latent classes there are while learning which entities belong to each class. We pursue a similar approach with a richer kind of latent variable -- latent features -- using a nonparametric Bayesian technique to simultaneously infer the number of features at the same time we learn which entities have each feature. The greater expressiveness of this approach allows us to improve link prediction on three datasets.

This paper presents a method for generating robust measurements by fusing the outputs of multiple elementary motion detector (EMD) circuits. The competitive feature tracker (CFT) algorithm is used to implement the EMD circuits. CFT measures motion by timing the movement of a feature across the visual field. Fusion is performed with leaky integrators. Each leaky integrator corresponds to a range of velocities and is stimulated whenever an EMD detects motion in the leaky integrator's velocity range. Thus the leaky integrator corresponding to the correct velocity receives the most stimulus. The other leaky integrators receive stimulus only from incorrect measurements. These incorrect stimuli are hence "voted out" and ignored. A proof-of-principle was obtained by implementing the algorithm in an optic flow sensor and exposing the sensor to moving texture stimulus. This sensor was then mounted onto a glider and placed in a control loop to cause the glider to steer away from a wall before collision.