different vlaues of inference

[masterblaster]

I have two networks which i have attached here. The first naive.xdsl is learnt from dataset using jsmile and myfile.xdsl is learnt from dataset using genie front end engine. Why the two networks are different even though they are learnt from a same dataset. The difference is that the four nodes are outside in case of structure learnt from genie but they contribute to belief when learnt from jsmile. And when i do inference in these two structure, the inference values are also different. Plz help me why this so.

[shooltz[BayesFusion]]

Can you post the data file you've used to learn the models? The snippet of Java code calling into jSMILE would also be helpful.

[masterblaster]

I think i found out the problem.

For example, if the node "ack" has two state 0 and 1 and for values not specified state "no1". The genie was quite intelligent to figure out that the node ack has states only 0 and 1 so there was state 0 and 1 only when learning structure from genie. But while learning structure from jsmile the node ack had states 0 and 1 plus state "no1".

Why it is so and can jsmile produce the same result as genie.

or my code is wrong. Below is the code i used to learn structure.

public void learn(String s)
{
NaiveBayes naive =new NaiveBayes();
DataSet D = new DataSet();
D.readFile(s);
Network net1 = new Network();
naive.setClassVariableId("type");
net1=naive.learn(D);
naive.setFeatureSelection(true);
net1.writeFile("naive.xdsl");
}