Restricting exchangeable nonparametric distributions
This work addresses a modeling limitation in nonparametric Bayesian methods for researchers and practitioners dealing with data where feature counts are not well-captured by standard priors, though it is incremental as it builds on existing models.
The authors tackled the problem that existing exchangeable nonparametric priors, like the Indian buffet process, may poorly model the distribution of features per data point in many tasks, and they proposed a new class of priors by restricting the domain of these models to allow specification of this distribution, achieving better performance on relevant datasets.
Distributions over exchangeable matrices with infinitely many columns, such as the Indian buffet process, are useful in constructing nonparametric latent variable models. However, the distribution implied by such models over the number of features exhibited by each data point may be poorly- suited for many modeling tasks. In this paper, we propose a class of exchangeable nonparametric priors obtained by restricting the domain of existing models. Such models allow us to specify the distribution over the number of features per data point, and can achieve better performance on data sets where the number of features is not well-modeled by the original distribution.