Indian buffet process
The Indian Buffet Process (IBP) is a generative model for random "feature allocations" (a "feature allocation" is analogous to a clustering except a given datum can belong to more than one cluster). So while the Chinese Restaurant Process describes a generative model for dividing N integers (customers) into K partitions (table assignments), the IBP describes a generative model for dividing N integers into K subsets, where each integer can occur in an arbitrary number of subsets. These subsets are known as "features" and the entire set is known as a "feature allocation". Note: the IBP has a more formal definition in probability theory where it is known as the marginalized distribution of a Beta process.
This concept has the prerequisites:
Core resources (read/watch one of the following)
→ The Indian Buffet Process: An Introduction and Review (2011)
Location: sections 3-4, though other sections provide useful background
Supplemental resources (the following are optional, but you may find them useful)
→ Nonparametric Bayesian Models (2009)
Location: Part 2 from 38:50
- The IBP linear Gaussian model is an application of the IBP to learning latent binary representations of observe data.
- create concept: shift + click on graph
- change concept title: shift + click on existing concept
- link together concepts: shift + click drag from one concept to another
- remove concept from graph: click on concept then press delete/backspace
- add associated content to concept: click the small circle that appears on the node when hovering over it
- other actions: use the icons in the upper right corner to optimize the graph placement, preview the graph, or download a json representation