In this article, a system for the research on probabilistic reasoning 
and learning from data base is reported. 
This system is based on a Bayesian network model,
and also it is expanded by neural networks for non-linear, continuous domain 
like many real world applications. Thus, this system can adapt to various
kind of data, and it can apply for more wide-spread applications.
Structural knowledge acquisition from large data base is recent promising 
research issue, and Bayesian network is also expected to abstract such 
knowledge in the structure of the network. 
For this purpose, learning structure of 
Bayesian network is studied actively. In order to evaluate variety of
structure selection algorithm, this system supports dynamical changing of
local tree structures. 
To handle practical large data base tractably, this system can connect to
major commercial, and non-commercial data base software in learning scheme
by JDBC. These features and GUI operations can help to find structural 
knowledge from large data bases.