BayesFusion Support Forum

Crash confirmed, we're looking into it now.

The messages you're getting show that your data is inconsistent with the parameters in the network. You have a bunch of deterministic nodes, and you're trying to input an impossible combination of evidence from the data file.

In the em file I generated CPTs randomly If you don't have any initial probability distribution for your model, just use EM's internal randomization (call em.SetRandomizeParameters(true), and optionally fix the random seed for the initial parameters with em.SetSeed before SetRandomizeParameters). D...

Your code will run fine when starting with ex7_TD2_10.xdsl. The log likelihood is around -990. If the model used is em_ex7_TD2_10_ED2_0.xdsl, the DSL_em::Learn will return DSL_ZERO_POTENTIAL, which is defined as -43. This is due to conflicting evidence in the dataset. Your code does not check for th...

We checked the log likelihood obtained after learning parameters with your files, and got values close to -1000 (in GeNIe and with a simple Python program.)

Maybe you did not create a proper matching between your network and data? Can you post your code fragment where you call EM?

Can you post your network and data file here, so we can reproduce the issue? Alternatively, send me a private message on the forum and attach the files (or provide a download link).

Please try again with the --upgrade --force-reinstall options:

SMILE 2.2 is now available. This version of the library supports new diagnosis algorithms. The libraries for C++, Python, Java, R and .NET can be downloaded from https://download.bayesfusion.com . We also maintain repositories for use with Maven and pip, see the download website for more details. Su...

The error is caused by C++ exception caught by code binding between core SMILE library and PySMILE. It's a memory issue. Are you running 32-bit or 64-bit Python?

I use the Learn parametres function

The 'Learn parameters' command only learns parameters, and works on the model structure already present. Are you referring to the output from the PC structure learning algorithm?

The min/max/avg are calculated over sensitivities of all parameters in node's CPT. The sensitivity algorithm calculates a separate sensitivity value for each of the CPT entry for each target state of each target node. You can view per-parameter sensitivities in the node properties/definition when se...

One more thing: if your calculate_stress function returns a list of values sampled from unknown distribution, then you can try to fit your samples to a metalog distribution. GeNIe/SMILE support Metalog since GeNIe version 4.0. To fit the data, use GeNIe's Tools|Metalog Builder window, or visit an on...

You can set node equation (as a string) with Network.set_node_equation. The actual formula depends on the type of the output from calculate_stress. To experiment with various node distributions using SMILE notation, you can use GeNIe's probability visualizer (available from the Tools menu), or go to...

Re: item 3, we will add an option to avoid zeros in the learned parameters. We already have similar functionality in the hybrid inference sampling algorithm. The replacement value for zero will be 1/N, where N is the numbers of records in the dataset.

Thanks for the long and detailed post. We will try to provide answers to all your questions in multiple posts. 1. We have added an entry in our issue tracker re: missing maxSearchTime for the PC learning. This is a simple change, and we should be able to implement it before the upcoming SMILE releas...