FAQ

  1. After I download the zip file, how do I run UnBBayes?
  2. I am running UnBBayes in my Mac and it does not seem to have a way of deleting nodes. Is there any secret way?
  3. How do I get UnBBayes' current source code?
  4. I downloaded the source code, but I am getting some errors when I try to compile it. How can I fix this?
  5. I want to use UnBBayes' API. Do you have any simple example I could look at?
  6. When do you think the manual will be available?
  7. Does it allow parallel processing?
  8. Does it run on a 32-bit or a 64-bit data bus?
  9. What kind of algorithm uses to perform inference?
  10. Can it handle continuous nodes, if so what kind (e.g. Gaussian, log-normal etc)?
  11. Does it perform parameter learning?
  12. Does it perform structure learning?
  13. Are dynamic networks supported?
  14. Does it do sensitivity analysis? If so, one way, two way, multidimensional?
  15. Does it provide the expected value of information?
  16. Can I input probabilities manually (e.g. not learned from data)?
After I download the zip file, how do I run UnBBayes?

You first have to unzip the file. After that just double click the jar file. If that does not work, run the following code in the prompt: >java -jar unbbayes-version.jar

[top]
I am running UnBBayes in my Mac and it does not seem to have a way of deleting nodes. Is there any secret way?

In Windows you just have to press the 'del' key, but in Mac you have to presse 'ctr'+'del' keys.

[top]
How do I get UnBBayes' current source code?

Although the CVS repository is active, we have stopped using it. Therefore, download UnBBayes current code you have to get it from our SVN repository as explained in the SVN source code menu at http://www.sourceforge.net/projects/UnBBayes.

[top]
I downloaded the source code, but I am getting some errors when I try to compile it. How can I fix this?

The easiest way to get UnBBayes' current version up and running is to download our source code in Eclipse using subversion plugin (http://subclipse.tigris.org/install.html). Once you download our source code, open the README.txt file and follow the instructions to install maven plugin in eclipse and make the project compile correctly (install the libraries into your local repository). We currently use Java 5. Due to some Mac compatibilities problems we did not migrate to Java 6 yet.

[top]
I want to use UnBBayes' API. Do you have any simple example I could look at?

To see how to use the exact inference go to our class in unbbayes.example.TextMode.java. There you will see the basic classes and methods needed to compile, add findings, propagate (update evidences), and get the marginal of a node.

[top]
When do you think the manual will be available?

It is ready, but it still needs to be reviewed carefully. The link is http://sourceforge.net/apps/mediawiki/unbbayes/index.php?title=Visit_the_Tutorial. If you find any error or if you have any suggestion related to our tutorial, please let us know.

[top]
Does it allow parallel processing?

We have not done anything related to that in our current work.

[top]
Does it run on a 32-bit or a 64-bit data bus?

I guess it depends on what Java compiler you use. I am pretty sure the releases we have available are for 32-bit. But you could compile the source code for a 64-bit.

[top]
What kind of algorithm uses to perform inference?

For exact we use Junction Tree and for approximate we have Likelihood Weighting and Gibbs Sampling.

[top]
Can it handle continuous nodes, if so what kind (e.g. Gaussian, log-normal etc)?

In our last release you are able to only represent the CLG (Gaussian), but we are working on implementing Junction Tree for CLG also. Besides that, we are also developing some distributions (like in Netica), including continuous, instead of just allowing the definition of CPTs (table entries) as it is done today. For the continuous nodes, the distribution will be discretized before performing inference (again, like in Netica).

[top]
Does it perform parameter learning?

Yes.

[top]
Does it perform structure learning?

Yes.

[top]
Are dynamic networks supported?

No. We currently support BN, ID, MSBN, OOBN, MEBN/PR-OWL, and partially CLG.

[top]
Does it do sensitivity analysis? If so, one way, two way, multidimensional?

No. Not in UnBBayes. We have another tool that might be helpful, UnBMiner. You should ask Prof. Dr. Marcelo Ladeira for that information.

[top]
Does it provide the expected value of information?

Not now. But it will as soon as we get the other distributions implemented.

[top]
Can I input probabilities manually (e.g. not learned from data)?

Yes.

[top]