Edit page

Using DPmem, we can create an infinite mixture model with arbitrary observation distributions. In the following model, the observation function is Gaussian:

(define class-distribution (DPmem 1.0 gensym))

(define object->class
  (mem (lambda (object) (class-distribution))))

(define class->gaussian-parameters
  (mem (lambda (c) (list  (gaussian 65 10) (uniform 0 8)))))

(define (observe object)
  (apply gaussian (class->gaussian-parameters (object->class object))))

(map observe '(tom dick harry bill fred))

See also:

References: