# miller2009

## Promise

Bayesian non-parametric method to infer both latent features and which entities have which feature.

Application : link prediction

## Delivery

* One author is a psychologist = accessible text to non experts.
* Good intro into latent feature approaches.
* Few words into
* class-based approaches : Stochastic block model
* Feature-based approaches : Like this one
* Generic method:
* Handles cases where entities belong to more than one categories / have more than one latent features.
* (Generalization of stochastic block models)
* Handles multiple relations between entities. (straightforward)
* Indian Buffet Process
* Implemented with Monte Carlo Markov Chains
* Interesting tricks inspired by bibliography

* No clear way to apply method in node addition (= matrix based, computationally non trivial)
* Implementation needs careful study (no code given)