Fahd: As pointed out earlier, the fact that you get redundancy across multiple inputs is what leads to "evidence" in the first place. For example, I would think that if both the 'light ON sensor' and the 'window open sensor' were jointly 'activated' that this is 'more evidence' that there is activity that occurs during 'daytime'. It might also tell me that the light source has higher probability of being 'from the open window' as opposed to 'from a light switch'.
I would suggest that you get yourself a copy of the book "Neural networks for pattern recognition" by Bishop as it covers the topic of Bayesian inference rather well. Then proceed to examine the following paper which discusses Bayesian networks rather well, Stephenson (2000): http://citeseer.nj.nec.com/cache/papers/cs/13350/ftp:zSzzSzftp.idiap.chz SzpubzSzreportszSz2000zSzrr00-03.pdf/stephenson00introduction.pdf Finally, you can build some intuition on how to apply Bayesian networks to your problem by examining some canonical examples (e.g., "What is the probability that person A is a smoker, given that person A was diagnosed with lung cancer?") . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
