Equation node: user-defined equation and post processing

[david_yang]

Hi!

I'm considering using equation nodes to establish a BN. However, the equations of some nodes involve functions that are not available in GENIE (e.g. error function). Can I define the function by myself using either GENIE or SMILE?
The second question is, after establishing the network, how can I get sample vales of one specific node so that I can plot priori and posteriori PDF and CDF.
The last question is, can I provide more than one evidence for one node?

Thanks in advance for your help!
David

[shooltz[BayesFusion]]

I'm considering using equation nodes to establish a BN. However, the equations of some nodes involve functions that are not available in GENIE (e.g. error function). Can I define the function by myself using either GENIE or SMILE?

You'll have to approximate the error function using the set of elementary functions and operators available in GeNIe.

The second question is, after establishing the network, how can I get sample vales of one specific node so that I can plot priori and posteriori PDF and CDF

In GeNIe, go to node's "Value" property page and select all the rows in the "Samples" grid. You can copy the numbers to the clipboard, then paste it into program you're using for PDF/CDF. Note that GeNIe can plot PDF/CDF at the same property page; you can also copy the chart for inclusion in other programs.

The last question is, can I provide more than one evidence for one node?

Try using GeNIe's "Virtual evidence" feature.