Lecture - 23 Reasoning with Bayes Networks

Lecture Series on Artificial Intelligence by Prof. P. Dasgupta, Department of Computer Science & Engineering, IIT Kharagpur. For more Courses visit http://nptel.iitm.ac.in

Category:

Education

Tags:

• Reasoning
• with
• Bayes
• Networks

[kalm77]

eventually, as you parse the inference chain, the T/F values disappear and you come to rely only on the original P(b) and P(e).

[kalm77]

excellent question! prof did not explain this in lec21. P(b), P(e) do NOT have parent (causal) variables. thus the probabilities are directly given as marginal probabilities (not conditional). However, in all the children (below the level of b, & e) we can and must take "conditional" probabilities. In these cond'l prob's, the b & e values only appear as T or F, bcs they are "given" as something that happened in earlier time! this is the essence of causality (time) in this particular belief net.

[mirce07]

(i'm not prof dasgupta) he considered that probabilyty after consulting some statistics(that's my thought)and perhaps in india this is the probabilyty for earthquake or burglary

[mistergmedina]

Hi Prof Dasgupta; I have a question : How do you obtain values for P(b), P(e), and why does this probably does not consider True/False states like as John and Mary according to "Belief Network Example" slide?

Thanks in advanced !