Latent feature models are widely used to decompose data into a small number
of components. Bayesian nonparametric variants of these models, which use the
Indian buffet process (IBP) as a prior over latent features, allow the number
of features to be determined from the data. We present a generalization of the
IBP, the distance dependent Indian buffet process (dd-IBP), for modeling
non-exchangeable data. It relies on a distance function defined between data
points, biasing nearby data to share more features. The choice of distance
function allows for many kinds of dependencies, including temporal or spatial.
Further, the original IBP is a special case of the dd-IBP. In this paper, we
develop the dd-IBP and theoretically characterize the distribution of how
features are shared between data. We derive a Markov chain Monte Carlo sampler
for a linear Gaussian model with a dd-IBP prior and study its performance on
several data sets for which exchangeability is not a reasonable assumption.