SmileX and Clustering algorithm

[jga]

Hello:

I am using SmileX with Excel. How could I use "Clustering" algorithm with SmileX?

SmileX manual says we can use Lauritzen, Henrion, Pearl, LSampling, SelfImportance, HeuristicImportance and Backsampling. All of them for BN. But not Clusterin.

Thanks very much

[shooltz[BayesFusion]]

I am using SmileX with Excel. How could I use "Clustering" algorithm with SmileX?

Lauritzen is clustering.

[jga]

Thanks, very much.

Then How can I explain the difference betwen the results with SmileX and Genie. I could compare the results obtained with Genie and SmileX with the same BN in two cases. With the first BN (very simple, with only 4 nodes) the results were very similar (insignificant difference), but with another BN with 30 nodes aprox. there are differences:

With GENIE: 0,764 - 0,253 - 0,979
With SmileX: 0,686 - 0,287 - 0,945

Is that normal?

Thaks

[shooltz[BayesFusion]]

Can you post your 30-node network?

[jga]

I posted it.

In rb.xdsl there are some evidences I use to compare Genie results and SmileX results. The target nodes are T1, T2, T3 and T4.

Thanks

[shooltz[BayesFusion]]

What exactly was the evidence you used to get the belief discrepancy? The model contains multiple cases.

[jga]

The first case. And the probabilities correspond to T1, T2 and T3.

[shooltz[BayesFusion]]

The first case. And the probabilities correspond to T1, T2 and T3.

Can't reproduce that with the following script (run from the command line with 'cscript test.vbs'):

Code: Select all

dim net
set net = CreateObject("SmileX.SmileXCtrl.1")
net.ReadFile "rb.xdsl"
net.SetEvidence "A", "Hombre"
net.SetEvidence "B", "E15_34"
net.SetEvidence "C", "Positivo"
net.SetEvidence "F", "Negativo"
net.SetEvidence "G", "Negativo"
net.SetEvidence "H", "Positivo"
net.SetEvidence "L", "Vida_cama_sillon"
net.SetEvidence "N", "Negativo"
net.SetEvidence "R", "Dificultas_Ronquidos"
net.SetEvidence "S", "Negativo"
net.SetEvidence "V", "Extension_en_respuesta_al_dolor"
net.SetEvidence "X", "Al_dolor"

dim res
res = net.UpdateBeliefs("Lauritzen")
Wscript.Echo "UpdateBeliefs status code:", res

dim b
b = net.GetBeliefs("T1")
Wscript.Echo "T1:", b(0), b(1)
b = net.GetBeliefs("T2")
Wscript.Echo "T2:", b(0), b(1)
b = net.GetBeliefs("T3")
Wscript.Echo "T3:", b(0), b(1)

The beliefs returned are in sync with GeNIe's values.

Code: Select all

T1: 0.235561961436078 0.764438038563923
T2: 0.021294296604007 0.978705703395993
T3: 0.746880495503305 0.253119504496695

Make sure your Excel/VB code checks the return values from the SetEvidence method (should be zero if the call succeeded).

[jga]

I apologise for this mistake, because the states of one node stored in excel must containe some strange character. The returns value of SetEvidence method is -1 in some cases.

Thanks very much and best regards