Can't nest rejection-query inside enumeration-query

[longouyang]

This might be an FYI more than a real bug. Non-working code:

(define (f n)
  (rejection-query
   (define x (uniform-draw (list n (+ n 1) (+ n 2) (+ n 3) (+ n 4))))
   x

   (>= x 5)))

(enumeration-query
 (define n (uniform-draw (list 1 2 3)))

 (define y (f n))

 y

 true)

Of course, we could just make f an enumeration-query and use (apply multinomial (f n))

[ngoodman]

Note: this is because rejectionSample in probjs is implemented in a way that makes a new trace object for each try, rather than staying within the enclosing trace. It could be done as an ordinary js helper function instead. It's not clear whether there are speed implications.

There is also some interaction with the (pending) fix to allow rejection to deal correctly with factor and directly conditioned ERPS.

[longouyang]

Noah, what is the status on this? Did the possible interactions get resolved?

[ngoodman]

i don’t think anyone went back to work on this / these issues.

-N

On Mar 28, 2014, at 6:31 PM, Long Ouyang [email protected] wrote:

Noah, what is the status on this? Did the possible interactions get resolved?

—
Reply to this email directly or view it on GitHub.