BayesFusion Support Forum

I believe your options for SMILE integration are: 1. if the backend runs on Node, write an add-on with N-API (https://nodejs.org/api/n-api.html#node-api). The add-on would link SMILE and perform the calculations. Please note that we have never done anything with Node add-ons, this suggestion is base...

Please refer to the 'Copyright notice' chapter in GeNIe manual, which is available from GeNIe's help menu, and also online (PDF and HTML formats) at https://support.bayesfusion.com/docs/

We have modified the hybrid inference algorithm to avoid the creation of the discretized network. SMILE 2.x performs the discretization on the fly and always takes into account the information available to reduce the size of the CPTs (constant nodes and evidence, including evidence set on discrete n...

Your version of R must be compatible with the R version we use to build rSMILE. As indicated in the rSMILE section at our download website, we currently use R 4.2.1. You need to update your R if you want to proceed with rSMILE 2.0.8.

Thanks for reporting this. I was unable to reproduce the exact scenario you described, but found other issues in copy/paste code. This issue gets high priority in our tracking system.

I added this to our tracking system. We'll change equation discretized state indexing to 0-based in the next service release.

We have modified the code for solving hybrid models the discretization is required. We no longer create a separate copy of the network with discrete nodes (which you could previously inspect with the Discretize command). Instead, the discretized CPTs are created on demand, and evidence in parents is...

Thanks for spotting the bug. We'll fix it in the next service release.

As a temporary workaround, please consider using the If() function, which supports state literals in GeNIe 4.0. For example:

Please update your GeNIe version to 4.0 - it's available at our download site (https://download.bayesfusion.com)

The default value of network's random seed is zero - you should get different sampling values each time you run inference.

To check the seed, go to network properties and select the 'Inference' page. Random seed is there (it is saved in the .xdsl file if you change it).

AHP is a completely different technique that some people compare to multi-attribute utility functions. I don’t believe there is any straightforward way to combine AHP with probabilities. Hence no straightforward combination with Bayesian networks.
I hope this helps.

See Tutorial3 in SMILE Wrappers Manual. The manual is available as HTML or PDF at https://support.bayesfusion.com/docs.

SMILE 2.0 is now available at https://download.bayesfusion.com . This version of the library supports metalog distributions and discrete nodes based on intervals or point values. SMILE fully supports Macs based on M series (Apple Silicon) ARM CPUs. The C++, Python and Java libraries are available as...

You can open the Value page in node properties, and copy the utilities from there. The numbers are copied in the text format, which can be pasted into any text editor or spreadsheet. If you want to run over multiple sets of evidence/decisions, and get utilities for each case written to a file, it ca...

The PC algorithm produces a 'pattern', which may include cycles and/or undirected or bidirected arcs. If your pattern is a DAG, you can proceed to EM by clicking on the 'Learn parameters' button in the PC patter editor.