Some suggestions reagrding Variable Elimination

Hi Arthur, Thank you for your comments. I shall try to answer them to the best of my abilities. 1. So the traditional method is inference by enumeration. So basically we are given the entire JPD (joint probability distribution) on set of variables and we need to say, for instance compute conditional probability given some evidence. How you would traditionally go about doing this is that you will first condition on the distribution to incorporate evidence e, which basically means that you will eliminate those records from your table which don't agree with your evidence and then normalize. After that you will marginalize on this new distribution to accommodate 'observation' part of the conditional probability. If we have 'n' variables and 'd' is the size of the largest domain, space complexity to store JPD is O(d^n). Time complexity is also O(d^n) because in the worst case we need to sum over all entries in the JPD. 2. I have given the algorithm in Section 5, which in essence is similar to pseudocode I guess? 3. Sorry I was updating my 'Applications' section when you wrote this. I hope that answers your question now. Thanks and regards, Ritika