Home

UnBBayes is a probabilistic network framework written in Java. It has both a GUI and an API with inference, sampling, learning and evaluation. It supports BN, ID, MSBN, OOBN, HBN, MEBN/PR-OWL, PRM, structure, parameter and incremental learning.

You can Browse Pages for more information.

MEBN - Edit

MSBN - Edit

BN - Inference

ID - Inference

Classification Performance Evaluation

OOBN - Edit

---

Project Admins:

• Shou Matsumoto
• Laecio Lima dos Santos
• Rommel Novaes Carvalho
• Marcelo Ladeira

Project Members:

• Cheol Young Park
• Diego Marques de Azevedo
• Guilherme Carvalho Torres
• Han Jun Yoon
• Laecio Lima dos Santos (admin)
• Marcelo Ladeira (admin)
• Paulo Cesar Costa
• Pedro da Costa Abreu Júnior
• Rafael Martins
• Rommel Novaes Carvalho (admin)
• Shou Matsumoto (admin)

License

---

Related projects

Developers in the open-source software community are also contributing to UnBBayes by publishing/distributing software that can be loaded to UnBBayes as plug-ins. Please, find below some examples of such featured open-source projects:

HML-UnBBayes Plugin

The HML (Human-aided Multi-Entity Bayesian Networks Learning) tool
enable users to create Multi-Entity Bayesian Networks (MEBN) from a
relational database (RDB), learning Local Probability Distributions
(LPD) of nodes from the data on the RDB.

EM-MEBN-training

This implements EM MEBN training in Java, using unbbayes libraries.
It was used in the Chantas et al 2018 paper published by ACM Journal of Computing for Cultural Heritage.
A port/wrapper which enables this program to be loaded as a plug-in of UnBBayes
is available in a separate fork.

[Bennila Mounira]

Does Unbbayes handle hybrid bayesian networks?
if yes, should we use the EM algorithm to learn the parameters knowing that we have incomplete data or should we use another algorithm ?

[Shou Matsumoto]

Unfortunately, the UnBBayes core and the plug-ins you can find under this sourceforge project page does not support hybrid Bayes nets.

However, there are some "3rd-party" plug-ins that add some features for handling hybrid Bayes nets. They are under the prognos sourceforge project (https://sourceforge.net/projects/prognos/).
You may need to build these plug-ins from source code, though, because the binary distributions are quite outdated already.

should we use the EM algorithm to learn the parameters knowing that we have incomplete data or should we use another algorithm ?

I believe you should use other implementations for parameter learning of hybrid Bayes nets.
I'm not aware of any parameter learning plug-in for hybrid Bayes nets in UnBBayes. The closest I know is the HML-UnBBayes Plugin (https://hml-unbbayes.github.io/), which can learn some parameters for Multi-Entity Bayesian Networks (although with some limitations).

For other/further questions, please use our discussion forum (https://sourceforge.net/p/unbbayes/discussion/). Thanks.

Best.