Markov Networks
We study this by comparing to Bayesian networks: Burglar, Earthquake, Alarm, He calls, She calls
– First, the values do not need to sum to one, that is, the table does not define a probability distribution. It only tells us that configurations with higher values are more likely. Second, there is no conditioning. It is proportional to the joint distribution of all the variables involved, as opposed to conditional distributions in CPDs.
– The tables are basically every combination of existing values: Let’s say potential function to capture relation between three vars A, B, CPDs where C is “soft XOR " of A and B, now we tabulize for all values of A, B, C and those that satisfy potential function will result in higher values in tables. e.g. 0 0 0 –> 10 0 1 0 –> 1 1 1 1 –> 1
–