Indian Buffet Process:
In the mathematical theory of probability, the Indian buffet process (IBP) is a stochastic process defining a probability distribution over sparse binary matrices with a finite number of rows and an infinite number of columns. (More on this process)
Question:
Consider an Indian Buffet Process (IBP) with parameter $\alpha$.
a) The primary definition of IBP is non-exchangeable. Explain the method that makes this process exchangeable.
b) Proof following properties:
- Adopting the above method makes the process exchangeable.
- Number of ones in each row follows $Poisson(\alpha)$.
- Expected total number of ones is $\alpha N$
What I've found so far:
According to the search results, the answer to all of these questions is written in the paper titled "Infinite Latent Feature Models and the Indian Buffet Process" by Griffiths and Ghahramani. However, I am not knowledgeable enough to understand what the paper says. I want someone to explain the answers. I have searched a lot but no explanation was provided anywhere! Everyone just cites the paper without any discussion. I would be very glad if someone provided something easier to understand as an answer to these questions.
