\begin{abstract}
Hybrid Intelligent Systems that incorporate the complementary features of
 expert systems and neural networks paradigms, have been studied by 
several researchers.
In this proposal, we introduce a novel Hybrid Intelligent Architecture (HIA)
that exploits the complementary features of rule based systems, statistical 
systems, 
and connectionist architectures. The proposed HIA augments a knowledge 
base system with a connectionist
 model and a statistical model to help the former to refine its domain
 knowledge. A novel feature of this HIA is the 
ability to refine the extracted domain knowledge and hence enhance the 
output decision taken by the proposed architecture.

The proposed HIA concepts were implemented on a real control problem and 
two classification problems. The experimental results showed that a 
connectionist network generated by the HIA generalizes better and 
faster than other connectionist architectures. Moreover, the HIA was 
able to refine the initial domain knowledge and hence enhances the 
output decisions of its connectionist architecture.
\end{abstract}