When is it safe to cancel a term when marginalizing over a distribution (to reduce the bulk of computation)?
Consider the example:
p(a,b,c,d,e) = p(c|a,b)p(a)p(b)p(d|a)p(e|d,b)
suppose we would like to find p(e=1|b=1) = p(e=1,b=1) / p(b=1)
|
When is it safe to cancel a term when marginalizing over a distribution (to reduce the bulk of computation)? Consider the example: p(a,b,c,d,e) = p(c|a,b)p(a)p(b)p(d|a)p(e|d,b) suppose we would like to find p(e=1|b=1) = p(e=1,b=1) / p(b=1) |
|||
|
|