Bayesian Network based on Neural Networks
(Bayonet)
(Awarded by Japanese Society of Artificial Intelligence, 1998.12.9)
[Japanese page]
[Status of the newest version]
We relase product version of Bayonet
.(September 2002)
We have developed two different versions, Bayonet-pilot and
Bayonet-real.
Download page of Bayonet-pilot and
its related page(in Japanese) has been opened.
Bayonet-real (current version) has been developed and demonstrated on RWC symposium already.
However, now we prepare documentation and we need Beta-testers.
If your are interested in, please contact to y.motomura@aist.go.jp.
(Complete documents in English will be described through discussion and questions.)
[Documents of Bayonet-real]
Manual (tentative)
Related papers
[Movie of Bayonet-real]
Video introduction (for rwc2001)[mpeg:50MB, Japanese speech]
Video introduction (for rwc2001, short version)[mpeg:2.7MB in Japanese speech]
[Features]
- Connectivity with SQL Database (Oracle, PostgreSQL and so on) and text(CSV) files.
- Learning conditional probabilities from the database with an efficient algorithm.
- Graph structure searching (Structure learning) with information criteria (MDL, AIC).
- Learning conditional probabilities with Neural networks and other learning models.
- Model construction using Wizard style operation.
- Network connectivity for client applications and programable command script and macros.
(all operation can be executed from TCP/IP connection as same as GUI.)
- Hugin, BayesBuilder compatible .net file generation (even the network include continuous, multi-dimensional variables).
- One step belief propagate calculation and monitoring. (Good for understanding the belief propagation algorithm.)
[Information of Bayonet]
Presentation in the international conference on softcomputing (AFSS2000) (in English)
Presentation on appling Bayesian network to user modeling domain(2001.1.10)
"Bayesian network learning system based on neural networks",
Proceeding of AFSS2000 (5pages: PS files) (in English)
Presentation in the annual conference of Japanse Society of AI (in Japanese)
Simple commentary in Japanese
Other information
Other links about Bayesian Networks
Previous version of Bayonet
Abstract of Bayesian Net(in Japanese)
Author's page
MOTOMURA Yoichi