Hilbert Space Embedding for Dirichlet Process Mixtures
This work addresses model selection and tractability issues in machine learning, though it appears incremental as it builds on existing stick-breaking and kernel methods.
The paper tackles the challenge of combining Bayesian nonparametrics with efficient kernel methods by proposing a Hilbert space embedding for Dirichlet Process mixture models, aiming to leverage the strengths of both approaches for improved inference and learning.
This paper proposes a Hilbert space embedding for Dirichlet Process mixture models via a stick-breaking construction of Sethuraman. Although Bayesian nonparametrics offers a powerful approach to construct a prior that avoids the need to specify the model size/complexity explicitly, an exact inference is often intractable. On the other hand, frequentist approaches such as kernel machines, which suffer from the model selection/comparison problems, often benefit from efficient learning algorithms. This paper discusses the possibility to combine the best of both worlds by using the Dirichlet Process mixture model as a case study.