Bayesiannetwork贝叶斯网络精品PPT课件
合集下载
相关主题
- 1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。
PY|X1,X2PX P 1,X X1 2,|X Y2 P YPX2P |Y X P 2P X1 X |Y 1PY PX2P |Y X P 2 Y |X1
Introduction
• Bayes theorem allows us to transform our prior probabilistic knowledge of an event into a more robust, posterior knowledge of its probability.
• P(Y|X)= PX |YPY PX
Introduction
• A train of consequences • Before the match has started, prior
possibilityP(Y); after observing the evidence X1, posterior probability P(Y|X1); new evidence X2 is acquired, then
Data
conditional probability tables
Hybrid Random Fields Bayes Networks
Jill 5/2/2013
Introduction
• For some mysterious reasons, • Bayes networks are often introduced in the
literature with examples concerning weather forecase: • Is it likely to be sunny on Sunday given the fact that it is raining on Saturday, and that my granny's back hurts?
d-seperation
• some terminology • head-to-tail at node B • tail-to tail at A
A B C A
B
C
A
C
• head-to-head at B
B
d-seperation
A
BБайду номын сангаас
C D
E
Z is observed: A,B are in Z; C, D or E are not in Z
• Prior confidence --------belief • Process--------belief propagation dynamics
causal relationships
statistical dependence between
Bayesian Networks
• DAG: Directed Acyclic Graph • CPT: Conditioanl Probability Tables
d-seperation
d-seperation is a sufficient condition for conditional independence
Markov Blankets
Markov Blanket of a node is the set consisting of its parents, children and spouses
Parameter Learning
• In order to fully specify the Bayesian network and thus fully represent the joint probability distribution, it is necessary to specify for each node X the probability distribution for X conditional upon X's parents
Markov Blankets
u1 un
z1
zn
Y1 Yn
feature A is independent of all other features given MB(A)={u1,...un,z1,...zn, Y1,...,Yn}
Markov Blankets
• For any value xi of Xi,
Introduction
• make prediction on whether your favorite football team will win, lose or draw tonight
• outcome of the match ---random variableY • Y=w, l, d • Rely on your original estimate of a certain
Directed Markov Assumption
• Each variable is independent of its nondescendents in the DAG given the values of its parents.
Directed Markov Assumption
prior possibility P(Y) ( rankings, recent history of their performance)
Introduction
• First half is over • The outcome of the first period may be
treated as a random variable X, the óbserved evidence' that influence your prediction of the final value of Y.
Advantage of BN: X 1,...aX l.ln oa .w.u,ro osne tolryeed sstta iime maan rttiene c gjoad ie nyrtedsl,w aistttc irvriebe a lua yttion lrnlbay n:sfo
small number of parameters
Introduction
• Bayes theorem allows us to transform our prior probabilistic knowledge of an event into a more robust, posterior knowledge of its probability.
• P(Y|X)= PX |YPY PX
Introduction
• A train of consequences • Before the match has started, prior
possibilityP(Y); after observing the evidence X1, posterior probability P(Y|X1); new evidence X2 is acquired, then
Data
conditional probability tables
Hybrid Random Fields Bayes Networks
Jill 5/2/2013
Introduction
• For some mysterious reasons, • Bayes networks are often introduced in the
literature with examples concerning weather forecase: • Is it likely to be sunny on Sunday given the fact that it is raining on Saturday, and that my granny's back hurts?
d-seperation
• some terminology • head-to-tail at node B • tail-to tail at A
A B C A
B
C
A
C
• head-to-head at B
B
d-seperation
A
BБайду номын сангаас
C D
E
Z is observed: A,B are in Z; C, D or E are not in Z
• Prior confidence --------belief • Process--------belief propagation dynamics
causal relationships
statistical dependence between
Bayesian Networks
• DAG: Directed Acyclic Graph • CPT: Conditioanl Probability Tables
d-seperation
d-seperation is a sufficient condition for conditional independence
Markov Blankets
Markov Blanket of a node is the set consisting of its parents, children and spouses
Parameter Learning
• In order to fully specify the Bayesian network and thus fully represent the joint probability distribution, it is necessary to specify for each node X the probability distribution for X conditional upon X's parents
Markov Blankets
u1 un
z1
zn
Y1 Yn
feature A is independent of all other features given MB(A)={u1,...un,z1,...zn, Y1,...,Yn}
Markov Blankets
• For any value xi of Xi,
Introduction
• make prediction on whether your favorite football team will win, lose or draw tonight
• outcome of the match ---random variableY • Y=w, l, d • Rely on your original estimate of a certain
Directed Markov Assumption
• Each variable is independent of its nondescendents in the DAG given the values of its parents.
Directed Markov Assumption
prior possibility P(Y) ( rankings, recent history of their performance)
Introduction
• First half is over • The outcome of the first period may be
treated as a random variable X, the óbserved evidence' that influence your prediction of the final value of Y.
Advantage of BN: X 1,...aX l.ln oa .w.u,ro osne tolryeed sstta iime maan rttiene c gjoad ie nyrtedsl,w aistttc irvriebe a lua yttion lrnlbay n:sfo
small number of parameters