DESCRIPTION:
Reasoning Systems
The concept of "Thinking Machines" has been around for a long time. Until now, they have failed to live up to expectations.
Programming languages have been around since the development of computers. They have been developed to allow logical structures in the form of well defined rules to be implemented for those computers. Computers and microcontrollers process the compiled programming languages into a series of steps that correspond to the strict rules defined for the application.
This is RULE PROCESSING, not reasoning!
Another form of "Thinking Machines" attempts to mimic the structure of the human brain. Artificial Neural Nets operate in this way. They are taught patterns to recognize.
This is PATTERN MATCHING, not reasoning! It is difficult or impossible to have a Neural Net explain "why" it made a decision or "why" it took a specific action.
For a system to "Reason" it must look at multiple situations simultaneously. Reasoning and Judgmental Decision-Making have long been considered human traits. Human reasoning is analog in nature. It is parallel, not sequential. It balances the pros and cons of various scenarios in order to select the best action or set of actions to perform. It considers how each action may impact other decisions or actions.
In many cases, reasoning is adaptive. It requires a constant re-evaluation of the importance of information. As information importance changes (as a factor of the age of the information, or distance between elements, or because of some other stimulating factor) the reasoning system must re-balance. Driving a car and treating a patient are adaptive processes.
Compsim's patented and patent-pending technology focuses on human-like reasoning that can be utilized to assist humans or can be embedded in devices or in software applications for autonomous or semi-autonomous operation where human-like expertise is needed.
Compsim's KEELŪ technology allows human reasoning to be captured in a model that executes in a human-like manner. It adapts to the changing importance of information. It balances a number of inter-related scenarios simultaneously.
When in the development phase, Compsim's technology allows the domain expert to "see the system reason". The development paradigm defines a new approach to "programming and documenting human reasoning".
Once embedded into devices, Compsim's Tools allow the reasoning process to be visualized. Like CAT scans (Computed Axial Tomography) in the medical space, Compsim's external tools allow the user to visualize the reasoning that is taking place "in the device". These are not "code analysis tools"; they are "reasoning analysis tools". Because KEEL based reasoning is explicit, the exact actions can be audited.
In some cases a small memory footprint is a key benefit. It is possible to embed KEEL technology in 8 bit microprocessors using only a few K-bytes of memory. This allows human-like reasoning to be deployed in low end sensors, cell phones and embedded devices; all the way up to high end microprocessors and mainframes.
KEEL technology is architecture independent. Systems can be built on a single stand-alone microprocessor or distributed across wide area networks. They can be loosely or tightly coupled reasoning systems.
KEEL technology, is packaged as KEEL Engines (or reasoning engines). These are functional subroutines that are automatically created as conventional programming language subroutines. This allows "reasoning extensions" to almost any existing product design without a significant re-engineering effort. But don't think you can look at the source code of the subroutines and understand the reasoning. It is the iterative processing of the data that generates the reasoning.
|
|
HOMEPAGE:
|
Click here to view
|
|
DEPARTMENT ID:
|
10-08149
(Last edited
on 2004/07/28 14:37:42 GMT-6
)
|
FACULTY/DEPARTMENT MEMBERS:
|
NAME
|
POSITION
|
EXPERTISE
|
PUBS
|
CTNS
|
|
Keeley, Helena
|
Other |
KEEL Technology; Cognitive Technology; Decision-Making Technologies; Decision Support Systems; Executive Management; Software Engineering; Robotics; Automation Systems |
0
|
0
|
Please note that the total of members' statistics may not equal the department/faculty
statistics due to joint authorship.
|
