الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 105 |  |  |
| | | from | 105 | | |
Abstract
Most developed learning algorithms are designed for environments in which all the relevant data is stored at single computer site. Therefore, previously developed learning algorithms for computations with such distributed databases require that all data stored in distributed locations must be transferred to a common site called central site and recompiled as one complete and local dataset before the construction can take place. The danger in this transfer is obvious if the data itself is innately sensitive. These databases cannot be moved to other network sites due to communication costs, security, data size, privacy, or data-ownership considerations. These databases together constitute the dataset from which information or patterns have to be extracted. An ecient system for computations with such distributed databases would work by doing as much work at local sites as possible and then communicating minimum required information (statistical summaries) between the sites to complete the computation tasks. This is much more ecient than transferring the complete databases to a single site, join these databases, and then execute algorithms with this data. The aim of this work is to give more power to the individual database sites and transferring minimum required information among the database sites. What has been attempted is to allow the individual database sites to perform calculations on their