Probability Kernel

Probability Kernel

by Gabin Paul Jacques Leroy -
Number of replies: 1

Hello,

I would like to have precisions about the definition of probability kernel. First, we have one discrete random variable X on a probability space (\Omega,F,\Bbb{P}) where we take the sigma field as the power set of \OmegaLet D_X denotes the alphabet of X. Then we consider another alphabet, associated to another discrete random variable Y on (\Omega,F,\Bbb{P}), that we call D_Y. So in the definition of the course, when it is written W(y|x), it means W(y|x)=\Bbb{P}( Y=y| X=x  ) and this is defined since both X and Y are defined on the same probability space. Is my interpretation correct ?


Bests,


Gabin

In reply to Gabin Paul Jacques Leroy

Re: Probability Kernel

by Thomas Weinberger -
Dear Gabin,

Yes, your intuition is correct! A probability kernel is simply a convenient representation of the conditional probability distribution of two discrete random variables. This representation is especially useful for studying Markov processes, as we can infer certain properties by studying matrix products and/or the spectrum of the probability kernels.

Best,
Thomas