probability distribution estimation from correlated samples

68 Views Asked by At

I am looking to solve the following estimation problem. Consider a blackbox where (given below) given an input X, its N observations are recorded as output.

These observations are denoted by $Y_1, \cdots, Y_N$. The observations are identically distributed, i.e. $Y_i \sim \mathcal{P}(Y|X) $. However, they are not independent.

I would like to know, how can I estimate $\mathcal{P}(Y|X)$ given these N observations. An exemplary solution for binary symbols would be great (i.e. $X,Y \in \{0,1\}$).

                X
                |
-----------------------------------
|                                 | 
|                                 |
|                                 |
|                                 |
-----------------------------------
  |      |     |      |        | 
 Y_1    Y_2   Y_3    Y_4      Y_N