Conference heron::euro_swas_ai

405.0. "Beologic Software" by FASDER::MTURNER (Mark Turner * DTN 425-3702 * MEL4) Tue Jan 28 1992 17:34

    Has anyone heard of this?  What are your impressions?
    
    
    						Thanks,
    
    
    						Mark
    
    	.............................................................
    
[forwarding deleted]
    
From:	NAME: Torben Brams @DMO             
	FUNC: DK Selected Accounts            
	TEL: (7)857 - 2434                    
<BRAMS.TORBEN@A1COPCLU@COPCLU@DMO>
To:	See Below
CC:	See Below
    
    It is our pleasure to invite you and any of your colleagues who may 
    be interested in innovative technology, to a presentation of the 
    BEOLOGIC software, which Bang & Olufsen (B&O) has developed in 
    cooperation with the Technical University of Denmark. 
    
    
    During this presentation you'll meet some of the people behind the 
    technology, see some of the application areas demonstrated, and have 
    the option of getting your questions answered.
    
    
    The event will take place in Nashua, New Hampshire 27. February as 
    two sessions at 10:00 AM and 2:00 PM in the BUSH conference room 
    (ZK02-1). We will repeat this event in Marlboro, Mass. 28. Feb. at 
    10:00 AM and 2:00 PM in the Luau Poi conference room (LMO pole I13).
    
    
    The BEOLOGIC technology is a revolutionary new and powerful tool, 
    which is perfectly suited for application areas where speed, space, 
    completeness and security is a priority.
    
	 * Diagnosis
	 * Simulation
	 * Configuration
	 * Problem solving
	 * Planning/Scheduling

    	 
    
    Digital Denmark has signed an exclusive distribution and marketing 
    agreement with B&O for BEOLOGIC on ULTRIX and VMS. This agreement 
    also gives Digital exclusive rights to distribute BEOLOGIC on UNIX, 
    once the market potential has been identified.
    
    
    If you and your organization needs innovative technology to help you 
    stay ahead of the competition, we believe that BEOLOGIC will interest 
    you and that you should take the time to attend this event. We would 
    appreciate, if you would confirm your attendance with Dorte Windfeld 
    @DMO (COPCLU::WINDFELD).
    
    
    Best regards,
    
    Torben Brams and John Alexander,
    Digital Equipment Corporation - Denmark
    
    
    
    

    
    
    
    BEOLOGIC, the Automated Reasoning Concept :
    
    
    Are you aware of how many different arrangements you can do on a 
    chess board with 8 queens, in such a way that two queens can never 
    strike each other ?
    
    It is easily figured out using BEOLOGIC - there are 92 solutions. 
    It is definitely not a trivial task in e.g. Prolog.
    
    
    BEOLOGIC is the name of a technology offering all known inference 
    methods implemented using a completely new approach. This 
    implementation allows a rule based inference engine to bypass the 
    traditional barriers set by the combinatorial explosion.
    
    
    In the following paragraphs some of the features of BEOLOGIC will 
    be highlighted. For your convenience we have made the 
    documentation available on the network. It can be copied from :
    
    	 COPCLU::BEO_GUI.PS	(Users Guide)
    	 COPCLU::BEO_REF.PS	(Reference Manual)
    
    
    
    
    MATHEMATICALLY PROVEN GENERALITY :
    
    All logical inferences are carried out by means of a few 
    fundamental array operations. All known inference techniques like 
    resolution, modus phonens or modus tollens are subsets of this 
    array approach, which is to be regarded as the generalized 
    inference technique. 
    
    The resolution principle, known from e.g. Prolog, is not used. As 
    an example, deduction is not based on a proof by contradiction, 
    but a straightforward simplified method.
    
    
    HIGH SPEED AND PARALLEL EXPANSION CAPABILITY :
    
    The logical transformations are based on a parallel search of 
    binary patterns. As the knowledgebase is not represented in a tree 
    graph (Prolog), the breadth-first, depth-first and backtracking 
    search is not performed.
    
    
    COMPACT KNOWLEDGE REPRESENTATION :
    
    The knowledge is represented as binary patterns in array form. The 
    size requirements of the knowledgebase are approximately 
    proportional to the number of rules, and independent of the number 
    of variables. Thus, there is no problem with exponential size 
    explosion; in a knowledgebase with N binary state variables, 
    information about all 2**N combinations is addressable from the 
    inference engine. The knowledgebase is of constant size during 
    state space search, i.e. no additional rules are added during 
    search and there is no garbage collection. The knowledgebase size 
    is fixed at compile time.
    
    
    GENERALIZED RULES :
    
    The rule representation is not limited to the well-known If-Then 
    structures in production rule systems or the clause form in 
    resolution based systems. BEOLOGIC supports constraint as well as 
    dynamic rules :
    
    -	 Constraint rules represent any propositional function or 
         relation between state variables.
    
    -	 Dynamic rules represent any causal relationship between state 
         variables.
    
    
    
    RULE INFORMATION CONTENTS :
    
    One BEOLOGIC rule, e.g. "A implies B" contains information of all 
    the combinations of the variables A and B. In practice an average 
    of seven normal production rules are required to represent a 
    BEOLOGIC rule like the one above.
    
    
    EFFICIENT AND CONSISTENT DEVELOPMENT FACILITIES :
    
    The BEOLOGIC Inference Engine makes it possible to identify and 
    manipulate inconsistent or superfluous knowledge. Therefore, rules 
    can be tested for contradictions, tautologies or redundancies.
    
    
    OPEN SYSTEM STRUCTURE :
    
    The tools are offered in a callable library, which makes the 
    system flexible in  use.
    
    
    FAST REAL-TIME EXECUTION :
    
    Due to the efficient knowledge base structure and the logical 
    transformations, a very high speed of processing is obtained.
    
    
    EXPLANATION FACILITY :
    
    The system supports a natural language reasoning feature. The 
    sentences derived by the BEOLOGIC Inference Engine are generated 
    from the system defined semantical elements and from the user 
    specified explanation text for rules and variables.
    
    The explanation text indices for variables can be included in the 
    loadable file for the BEOLOGIC Inference Engine. This feature 
    secures consistency in variable processing in an application 
    program.
    
    
    
    
    INFERENCE OPTIONS :
    
    All known inference methods are subsets of the generalized method 
    as implemented, e.g. resolution principle, backward and forward 
    chaining etc.
    
    
    Deduction or normal forward analysis, informs the users of the 
    consequences resulting from given stimuli. In addition, it 
    explains how a given state is determined. 
    
    Abduction (diagnosis) informs the user which conditional state 
    implies a certain output. 
    
    Derived rule finds the relationship between different system 
    attributes.
    
    Theorem proving provides redundancy checking
    
    Solutions, labeling or enumeration finds all possible solutions to 
    a given set of facts and rules.
    

    Yes, we did an evaluation of it and almost started a project using it;
    we are still actively exploring opportunities where it could be used.
    Chris (CLAVIS::)WILD did most of the work.
    
    Cheers
    Steve

Mark,
	It is worth while going, especially if you get an opportunity to discuss
details of the implementation. Some features.-

	Deduction, Abduction, Induction, Theorem proving, Derived rules
	C based, multi-platform support (e.g. VAX and PCs)
	KB size Directly proportional to number of rules and indeptendent of
				the number of state variables.
	Can be embedded in conventional code.

	Chris Wild can give you the Maths.

Joe.