文档介绍：In
Pro ceedings of the T en th Conference on Uncertain t y in Arti
cial In telligence
Generating Ba y w orks from Probabilit y Logic Kno wledge
Bases
P eter Hadda wy
Departmen t of Electrical Engineering puter Science
Univ ersit y of Wisconsin
Milw auk ee
PO Bo x
Milw auk ee
WI
haddawy
cs
uwm
e du
The approac hkno wn as kno wledge
based mo del con
Abstract
struction W ellman et al
has attempted to ad
dress these t w o limitations
The idea is to represen t
W e presen t a metho d for dynamically gen
a class w orks with a kno wledge base of proba
erating Ba y w orks from kno wledge
tences con taining quan ti
ed v ariables and to
bilit ysen
bases consisting of
rst
order probabilit y
instan tiate a subset of these sen tences to generate a
logic sen tences
W e presen t a subset of proba
net w ork for solving a giv en inference problem
The
bilit y logic su
cien t for represen ting the class
w ork is a subset of the domain mo del
of Ba y w orks with discrete
v alued
represen ted b y the collection of sen tences in the kno wl
no des
W e imp ose constrain ts on the form of
edge base
Sev eral concrete approac hes to ac hieving
the sen tences that guaran tee that the kno wl
this functionalit yha v e b een prop osed P o ole
edge base con tains all the probabilistic infor
Go