Inputs: alarm.bif(Bayesian Inference file) which has relation of components gold_alarm.bif which ahs the correct probabilities records.txt file that has records of 11000 patients but each records missing one entry BayesNet.png which shows the newtork in pictoral form
Output: solved_alarm.bif which matches golden_alarm.bif
Algorithm used: Expectation-Maximization
As we don't have proper information we first expect the probability and then try to maximize our expectation I used laplace smoothing Implemeted an epsilon of 0.0005 for computing convergence.
command: python main.py