Elbrus Developers Group

Group Overview | People | Partners | Group News
| UM-Kailash | HORUS | OpenNet | Futures Gateway

SourceForge.net Logo

UM
Cognitive
Processing
Environment


Go to the PROJECT
thin bar graphic spacer

What is UM ?

UM (an acronym for many descriptives of the system, one of which is "Unified Methodology") is a systematic integration of different mechanisms, algorithms and methods for automated and interactive knowledge acquisition, inference and reasoning, learning, administration (management) and synthesis (innovation).

The project is an ambitious and comprehensive undertaking to provide a practical and flexible programming and application environment for such knowledge discovery and management goals. The project builds upon more than two decades of individual and team research in both the theory and applications of artificial intelligence, neural networks, genetic algorithms, self-organization, complexity theory, nonlinear processing, parallel distributed processing, and other specializations.

UM is a combination of many things, some of which can be used with "lives of their own" in terms of application and re-use by being Open Source software. It is a web-based environment for individuals and communities to use in the context of building, maintaining, contributing to or simply using an INcyclopedia, where the term "INcyclopedia" is taken to mean an internet-based, interactive encyclopedia on one or more topics. It is a back-end engine for performing pattern matching, recognition, classification and routing on data objects of different composition and type. It is a knowledge base of programmer-supplied, user-supplied, and automatically-generated rules and rule-like expressions for performing deductive, inductive and abductive reasoning tasks. It is also an API for others to use in making adaptations to their specific applications.

There are several components within the UM software space, all of which are presently in the public domain or under various Open Source licenses such as GNU General and GNU Library (Lesser) licenses. However, UM software is being designed and implemented to be usable and to have the ability for interactions with non-open systems and products as well, in order that it can become in the future a possible standard model and software engine/library for many different users.

One of the underlying components of the main UM framework, that which provides its workspace and the medium for the input, editing, exchange and testing of content and the learning of new patterns and rules, is the Tiki CMS. UM uses a component like Tiki, and databases such as MySQL, Oracle, SQL Server, and SAP, in the manner that such CMS applications - Wiki, Tiki or otherwise designed, use MySQL and different databases for their operational and data requirements.

Another component to UM is the HORUS engine, also Open Source, that provides the construction path and the maintenance for the Syntopicon database, a thesaurus-like construction that uses relational definitions and associations to build up a knowledge base of topics, themas and also cross-references and similarity nets among concepts, articles and other documents, images and different URL objects.

UM is in the early period of its implementation, having progressed through many forms on the theoretical stage and with several parts having been built experimentally during the course of research projects. Most of the older prior code is not being used directly in the new project as too much has changed in the way of compilers and operating systems. There are chunks and pieces of code ranging from Common LISP to Visual C++ and now the emphasis from a software development perspective has turned to Java and PHP as two different but complimentary and suitable languages and programming environments for the future work.

One of the central and highly ambitious modules within UM, intended to be server-based but allowing future distribution over networks of voluntarily client machines, is the BINAR Colony. A BINAR is a "bicameral nonlinear associative recognizer", which translates into being a software agent that can self-modify its own algorithm based upon a genetic-like code that is dependent not only upon its own structure but the behavior and changes within associated or affiliated BINARs in the system. This makes the BINAR behave in some respects like a cellular automata.

The UM operating "umbrella" program prevents the BINARs from getting "out of hand" in any sense, such as running an excessive number of agents, code that is essentially "no-op", or code that could consume disk, CPU or network resources. It is envisioned that as UM evolves, there will be certain "genotypes" of BINARs that emerge, partially through direct programming and partially through self-organized evolution, that are more useful and practical for certain types of applications. As these functions and their APIs become evaluated by a community of UM users, through ranking and voting that is already built into the CDE (e.g., the Tiki base environment), select BINARs can be packaged and released as Open Source libraries for general purpose application. The architecture is being defined in such a way that the BINAR software agents can be "neutered" and used as non-reproducing, non-self-modifying "ordinary code" that will remain static, except at the hands of modification by other programmers (in the typical Open Source fashion).

Additional information on the BINAR model and theoretical basis is provided in several published papers that will be made available soon. Presently one can explore some things at Ye Old Public Library

Another component of UM is a package developed over a 2-year period called ACEserver and it is an ETL (extract-transfer-load) or "data mover" application designed originally to pull large blocks of updated rows from different relational databases and to evaluate the contents and then store them into data warehouses. ACEserver has a novel graph-theoretic base for handling the independent processes of moving data rows from a given source (e.g., Oracle) to a given target (e.g., Teradata) and was designed to enable faster throughput and rows/sec data movement performance.

ACEserver was written in Visual C++ for a Windows 2000 platform but was designed with platform and OS independence in mind. Almost all components need to be re-coded but the package can be configured to fit several kinds of ODBC-compliant databases for input sources and output targets. The value of ACEserver within UM is that a very large number of rows from public-access or license-accessible databases of media, news, articles, and other matter can be examined for unique or changing content, then stored and archived, or discarded, or sent to specific users for further action. ACEserver offers a tool that can be very useful for large encyclopedia-type content and knowledge management systems.

Papers, Presentations, Memos, Notes, Fragments
Summary (@ 144pp) on several interrelated and interdependent contributing technologies and inventions feeding into UM and KAILASH in general

Contact Information
Project Administrator: Martin Dudziak Forte Horizons LLC, prime contributor & financial sponsor


Group Overview | People | Partners | Group News
| UM-Kailash | HORUS | OpenNet | Futures Gateway