LAWS@SRI-AI.ARPA (05/06/85)
From: AIList Moderator Kenneth Laws <AIList-REQUEST@SRI-AI> AIList Digest Monday, 6 May 1985 Volume 3 : Issue 58 Today's Topics: Administrivia - Request for Help, Seminars - Database Theory and Equations (MIT) & Processing Uncertain Knowledge (UToronto) & Compact LISP Machine (SU) & How Processes Learn (CMU) & Space Modeling for Robot Navigation (SU) & Pictorial Explanations (UPenn) & Distributed Knowledge-Based Learning (USCarolina) & DART: An Automated Diagnostician (SU) & Abstraction and Classification in NIKL (MIT) & Evidential Reasoning in Semantic Networks (BBN) ---------------------------------------------------------------------- Date: Sun 5 May 85 16:09:01-PDT From: Ken Laws <Laws@SRI-AI.ARPA> Reply-to: AIList-Request@SRI-AI.ARPA Subject: Seminar Abstracts I fell behind on the seminar notices when I took off for DC early this month. Rather than just ignore all the talks, I've decided to "read them into the record" now. It will take three issues to hold all the post-dated notices -- just ignore them if abstracts aren't your thing. I find that I'm spending too much time gathering and editing seminar notices, so I'm going to cut back. I will continue to forward the Stanford notices, but I'd like to get some volunteers to edit and forward notices from CMU, CSLI, MIT-MC, Rutgers, and UTexas-20. Other contributors can help me by providing meaningful Subject lines for your messages. I have been cleaning up many of the subject lines as an aid to sorting the messages and to simplify construction of the Today's Topics header. It takes a fair amount of effort to find a concise summary of an author's message, and I would appreciate it if contributors would make an effort to provide the summary. If anyone would like to split off a linguistics/psychology list, I'll provide the necessary assistance. I like reading the messages, but I'm having trouble moderating such discussions and deciding which seminar notices to include. I think that a separate discussion list is the best solution, but I'm willing to consider some kind of joint moderation of the AIList message stream. AIList is a fun hobby, but I'd like to have a little more free time for other activities. Thanks to everyone for helping to make AIList such a success. -- Ken Laws ------------------------------ Date: 5 Apr 1985 1229-EST From: ALR at MIT-XX.ARPA Subject: Seminar - Database Theory and Equations (MIT) [Forwarded from the MIT bboard by SASW@MIT-MC.] DATE: TUESDAY, APRIL 9, 1985 PLACE: NE43-512A Database Theory and Computing with Equations Paris C. Kanellakis MIT Databases and equational theorem proving are well developed and seemingly unrelated areas of Computer Science research. We provide two natural links between these fields and demonstrate how equational theorem proving can provide useful techniques and tools for a variety of database tasks. Our first application is a novel way of formulating database constraints (dependencies) using equations. Dependency implication, a central computational problem in database theory, is transformed into reasoning about equations. Mathematical techniques from universal algebra provide new proof procedures and better lower bounds for database dependency implication. Our second application demonstrates that the uniform word problem for lattices is equivalent to implication of dependencies expressing transitive closure together with functional dependencies, (functional dependencies were the first and most widely studied database constraints). This natural generalization of functional dependencies, which is not expressible using conventional database theory formulations, has an efficient decision procedure and a natural inference system. This is joint work with Stavros S. Cosmadakis. HOST: Prof. Guttag ------------------------------ Date: Wed, 10 Apr 85 13:04:45 est From: Voula Vanneli <voula%toronto.csnet@csnet-relay.arpa> Subject: Seminar - Processing Uncertain Knowledge (UToronto) UNIVERSITY OF TORONTO DEPARTMENT OF COMPUTER SCIENCE (SF = Sandford Fleming Building, 10 King's College Rd.) ARTIFICAL INTELLIGENCE SEMINAR - Monday, April 15, 11 a.m., SF 1105 Harry Stephanou Texas Processing Uncertain Knowledge The methodology described in this talk is motivated by the need to design knowledge based systems for applications that involve: (1) subjective and/or incomplete knowledge contributed by multiple domain experts, and (ii) inaccurate and/or incomplete data collected from different measure- ments. The talk consists of two parts. In the first part, we present a quantitative criterion for measuring the effectiveness of the consensus obtained by pooling evidence from two knowledge sources. After a brief review of the Dempster-Shafer theory of evidence, we intro- duce a set-theoretic generalization of entropy. We then prove that the pooling of evidence by Dempster's rule decreases the total entropy of the sources, and therefore focuses their knowledge. In the second part of the talk, we present a fuzzy classification algorithm that can utilize a limited number of unreliable training samples, or prototypes. We then pro- pose the extension of this algorithm to a reasoning by anal- ogy scheme in which decisions are based on the "similarity" of the observed evidence to prototypical situations stored in the knowledge base. The measure of similarity relies on a set-theoretic generalization of cross-entropy. ------------------------------ Date: Mon 8 Apr 85 14:21:11-PST From: Susan Gere <M.SUSAN@SU-SIERRA.ARPA> Subject: Seminar - Compact LISP Machine (SU) EE380/CS310 Seminar on Computer Systems Title: Compact LISP Machine Speaker: Steven D. Krueger Texas Instruments, Dallas Time: Wednesday, April 10 at 4:15 p.m. Place: Terman Auditorium The Compact LISP Machine (CLM) development program is the first of several Defense Advanced Research Projects Agency (DARPA) programs intended to provide embedded symbolic computing capabilities for government applications. As one of many contracts funded under the Strategic Computing Program, the CLM will provide a symbolic computer capability for insertion of artificial intelligence (AI) and robotics technology in a wide range of applications. The heart of the CLM system is a high speed 32-bit VLSI LISP processor chip, built using high speed CMOS technology. It is based on the architecture of the Explorer LISP machine from Texas Instruments, which is based on the CADR LISP Machine from MIT and and LISP Machines Incorporated (LMI). The CLM system consists of four types of modules: Processor, Cache/Mapper, 4MB Memory, and Bus Interface. The Processor, Memory, and Bus Interface modules communicate over a high-performance 32-bit multi-master NuBus(TM) system bus. Some motivation will be given for adopting a special architecture for symbolic processing. Then the basic architecture of the CLM processor and Explorer processor will be reviewed. The NuBus system bus will be described, and the CLM system modules will be described. ------------------------------ Date: 4 April 1985 1155-EST From: Theona Stefanis@CMU-CS-A Subject: Seminar - How Processes Learn (CMU) Name: PS Seminar - J. Misra, The University of Texas/Austin Date: 10 April (Wed.) Time: 3:00-4:30 Place: 4605 WeH Topic: How Processes Learn A key feature of distributed systems is that each component process has access to its own state but not to the states of other processes. Any nontrivial distributed algorithm requires that some processes learn about the states of others. To study such issues, we introduce the notion of isomorphism among computations: two computations are isomorphic with respect to a process if the process can't tell them apart. We show that isomorphisms can be used to define and study learning by processes. We give a precise characterization of minimum information flow for achieving certain desired goals. As an example we show that there is no algorithm to detect termination of an underlying computation using a bounded number of overhead messages. This talk assumes no previous background in distributed systems. ------------------------------ Date: 09 Apr 85 1321 PST From: Marianne Siroker <MAS@SU-AI.ARPA> Subject: Seminar - Space Modeling for Robot Navigation (SU) SPECIAL ROBOTICS SEMINAR On Wednesday, April 10, Raja Chatila from France will speak on Consistent Space Modeling for Mobile Robot Navigation Time: 4:15 PM Place: Rm 352 MJH In order to understand its environment, a mobile robot should be able to model it consistently, and to locate itself correctly. One major difficulty to be solved is the inaccuracies introduced by the sensors. The presented approach to cope with this problem relies on general principles to deal with uncertainties: the use of multisensory system, favoring of the data collected by the more accurate sensor in a given situation, averaging of different but consistent measurements of the same entity weighted with their associated uncertainties, and a methodology enabling a mobile robot to define its own reference landmarks while exploring and modeling its environment. These ideas are presented together with an example of their application on the mobile robot HILARE. ------------------------------ Date: Mon, 15 Apr 85 14:15 EST From: Tim Finin <Tim%upenn.csnet@csnet-relay.arpa> Subject: Seminar - Pictorial Explanations (UPenn) AUTOMATING THE CREATION OF PICTORIAL EXPLANATIONS Steve Feiner (Brown Univ) 3pm Monday April 15th, 337 Towne Building Computer and Information Science, University of Pennsylvania The APEX [Automated Pictorial EXplanations] project has as its long-term goal the realtime computer generation of effective pictorial and textual explanations. Our current research has concentrated on the automated creation of pictures that depict the performance of physical actions, such as turning or pushing, on objects. We are constructing a test-bed system that generates pictures of actions performed by a problem solver. Our system supports rules for determining automatically the objects to be shown in a picture, the style and level of detail with which they should be rendered, the method by which the action itself should be indicated, and the picture's viewing specifications. A picture crystallizes about a small set of objects inferred from the nature of the action being depicted. Additional objects and detail are added when it is determined that they help disambiguate an object from others with which it may be confused. ------------------------------ Date: Wed, 17 Apr 85 12:48 EST From: Huhns <huhns%scarolina.csnet@csnet-relay.arpa> Subject: Seminar - Distributed Knowledge-Based Learning (USCarolina) CENTER FOR MACHINE INTELLIGENCE University of South Carolina A DISTRIBUTED KNOWLEDGE-BASED LEARNING SYSTEM FOR INFORMATION RETRIEVAL Speaker: Uttam Mukhopadhyay Date: 3:00 p.m. Wednesday, April 17, 1985 Location: Room 230, Engineering Building MINDS (Multiple Intelligent Node Document Servers) is a distributed system of knowledge-based query engines for efficiently retrieving multimedia documents in an office environment of distributed workstations. By learning document distribution patterns, as well as user interests and preferences during system usage, it customizes document retrievals for each user. In this talk we discuss the implementation of a two-layer learning testbed for studying plausible heuristics. In the simulated environment, document distribution patterns (object-level concepts) used by the query engine are learned at the lower level with the help of heuristics for assigning credit and recommending adjustments; these heuristics are incrementally refined at the upper level with the help of meta-heuristics. ------------------------------ Date: Thu 25 Apr 85 16:35:45-PST From: Elliott Levinthal <LEVINTHAL@SU-SIERRA.ARPA> Subject: Seminar - DART: An Automated Diagnostician (SU) Professor Michael Genesereth will be the featured speaker at the Seminar on May 1st. Time is 2:15, in Terman Room 217. "DART: An Automated Diagnostician for Equipment Failures" Michael R. Genesereth Logic Group Knowledge Systems Laboratory, Stanford This talk describes a device-independent diagnostic program called DART. DART differs from previous approaches to diagnosis taken in the Artificial Intelligence community in that it works directly from design descriptions rather than MYCIN-like symptom-fault rules. DART differs from previous approaches to diagnosis taken in the design-automation community in that it is more general and in many cases more efficient. DART uses a device-independent language for describing devices and a device-independent inference procedure for diagnosis. The resulting generality allows it to be applied to a wide class of devices ranging from digital logic to nuclear reactors. Although this generally engenders some computational overhead on small problems, it facilitates the use of multiple design descriptions and thereby makes possible combinatoric savings that more than offsets this overhead on problems of realistic size. ------------------------------ Date: 26 Apr 1985 16:05 EST (Fri) From: "Daniel S. Weld" <WELD%MIT-OZ@MIT-MC.ARPA> Subject: Seminar - Abstraction and Classification in NIKL (MIT) ABSTRACTION AND CLASSIFICATION IN A HYBRID REPRESENTATION SYSTEM Marc Vilain BBN Laboratories Cambridge, MA Hybrid architectures have been used in several recent knowledge representation systems. In this talk, I will explore several hybrid representation architectures, focusing particularly on the architecture of the KL-TWO system. KL-TWO is built around a propositional reasoner called PENNI (a descendant of RUP) and a terminological reasoner called NIKL (a descendant of KL-ONE). I will show how NIKL can be interfaced to PENNI so as to augment PENNI's propositional language with a limited form of quantification. This interface relies crucially on two operations that follow naturally from the KL-TWO architecture: abstraction and classification. I will describe these operations, and discuss how their generality might extend beyond the scope of KL-TWO. Tuesday, April 30; 4:00pm; 8th Floor Playroom ------------------------------ Date: 24 Apr 1985 12:24-EST From: AHAAS at BBNG.ARPA Subject: Seminar - Evidential Reasoning in Semantic Networks (BBN) [Forwarded from the MIT bboard by SASW@MIT-MC.] Another BBN AI Seminar: Lokendra Shastri of U. of Rochester will talk on Friday April 26 in room 5/143 (near the travel office). Evidential Reasoning in Semantic Networks A Formal Theory and its Parallel Implementation The talk presents an evidential approach to knowledge representation and inference wherein the principle of maximum entropy is applied to deal with uncertainty and incompleteness. It focuses on a representation language that is an evidential extension of semantic networks, and develops a formal theory of inheritance and recognition within this language. The theory applies to a limited, but interesting, class of inheritance and recognition problems, including those that involve exceptions, multiple hierarchies, and conflicting information. The resulting theory may be implemented as an interpreter-free, massively parallel network made up of highly interconnected but extremely simple computing elements. The network can solve inheritance and recognition problems in time proportional to the depth of the conceptual hierarchy. ------------------------------ End of AIList Digest ********************