Soft Evidence Implementation

[khijuni]

Hi, all!

First of all, I really appreciate all the efforts that SMILE & Genie related personnel made. It is really wonderful BNs framework.

I currently use SMILE for building my web application. But I got stuck with a problem. On genie, I found a soft evidence functionality and want to apply it to my web application. I am using C++ SMILE API. If you can share some tutorial code with me, it will be great. I tried to find related article. But I couldn't.

For my hard evidence, I used the following code (just a partial part)

theNet.UpdateBeliefs();

string name;
int evidence_number = 0 ;
name = evidence_input[i+1] + "_Facility_" + evidence_input[i] +"_Report" ; // evidence input arrays are inputs from external txt file and ignore i variable
set_evidence= theNet.FindNode((char *)(name.c_str()));
theNet.GetNode(set_evidence)->Value()->SetEvidence(evidence_number);

What I want to do in place of my old code is that I want to find a node then set it as soft evidence. If the node has updated probability from other nodes, I want to use the updated probability as soft evidence as I can do it in Genie.

Thanks in advance!!

[shooltz]

SMILE's functionality for soft evidence is exposed through methods declared in DSL_nodeValue class:

Code: Select all

int GetSoftEvidence(std::vector<double> &evidence) const;
int SetSoftEvidence(const std::vector<double> &evidence);
int IsSoftEvidence() const;

Two first methods are virtual and actual implementation is in DSL_beliefVector.

The parameter passed to/from Get/SetSoftEvidence is the vector of probabilitites with the size equal to the node's outcome count. If you have a ternary node, your code could look like this:

Code: Select all

DSL_node *pNode = net.GetNode(handle);
pNode->GetValue()->SetSoftEvidence();
net.UpdateBeliefs();

Finally, use DSL_nodeValue::ClearEvidence to remove the soft evidence from the node.

[khijuni]

Thanks for your reply.

I have finished the coding and share with others, as I got great help from you ^^.

string facility_soft_evidence;
int credibility_update;
DSL_node *pNode;
DSL_nodeValue *pNodeValue;

facility_soft_evidence = "node_name" ;
credibility_update= theNet.FindNode((char *)(facility_soft_evidence.c_str()));
pNode = theNet.GetNode(credibility_update);
pNodeValue = pNode -> Value() ;

DSL_sysCoordinates theCoordinates_soft(*theNet.GetNode(credibility_update)->Value());
theCoordinates_soft[0] = 0;
theCoordinates_soft.GoToCurrentPosition();
vector<double> marks (2);
for (int y = 0 ; y <2 ; y++ ) {
marks[y] = theCoordinates_soft.UncheckedValue();;
theCoordinates_soft.Next();
}
pNodeValue -> SetSoftEvidence(marks);

theNet.UpdateBeliefs();

P.S. I am not sure this is efficient code. But SetSoftEvidence function seems to be tricky, as it requires vector as parameters. If anybody knows better way, please share with us. Thanks!