The First

Classics (Collection 2)

In this paper, we give an overview of our casebased reasoning program, HYPO, which operates in the field of trade secret law. We discuss key ingredients of case-based reasoning, in general, and the correspondence of these to elements of HYPO. We conclude with an extended example of HYPO working through a hypothetical trade secrets case, patterned after an actual case. As anyone, who has ever endured an adversarial "Socratic" dialogue with a forcefully probing judge, client, or law school professor, knows, being able to organize and marshal, one's cases is key to prevailing in argument. It is certainly not enough to cite maxims, sections of statute, or even simply list the relevant cases.

25 How to See a Simple World: An Exegesis of Some Computer Programs for Scene Analysis

Classics (Collection 2)

Department of Computer Science University of British Columbia This is not a comprehensive survey of machine vision which, in its broadest sense, includes all computer programs that process pictures. Restricting attention to scene analysis programs that interpret line data as polyhedral scenes makes it possible to examine those programs in depth, comment on revealing mistakes, explore the interrelationships and exhibit the thematic development of the field. Starting with Roberts' seminal work which established the paradigm, there has been an evolutionary succession of programs and proposals each approaching the problem with a different emphasis. In addition to Roberts' program this paper expounds in detail work done by Guzman, Falk, Huffman, Clowes, Mackworth, and Waltz. These programs are presented, compared, contrasted and, sometimes, criticized in order to exhibit the development of a variety of themes including the representation of the picture-formation process, segmentation, support, occlusion, lighting, the scene description, picture cues and models of the world.

13 Decision Trees and Multi-Valued Attributes

Classics (Collection 2)

Common induction systems that construct decision-trees have been reported to operate unsatisfactorily when there are attributes with varying numbers of discrete possible values. This paper highlights the deficiency in the evaluation of the relevance of attributes and examines a proposed solution. An alternative method of selecting an attribute is introduced which permits the use of redundant attributes. Results of experiments on two tasks using the various selection criteria are reported. As knowledge-based expert systems play an increasingly important role in artificial intelligence, more attention is being paid to the problem of acquiring the knowledge needed to build them.

Blank Page

Classics (Collection 2)

Feb 1481 card 1 of 1 BLANK PAGE Page 2 1. INTRODUCTION The impetus for work in Intelligent CAI has two major sources: firstly, the practical aim of producing teaching systems which are truly adaptive to the needs of the student and secondly the "theoretical" interest involved in formulating these activities as algorithms. It has been argued by Hartley and Sleeman. A numoer of sy.tems have been implemented during the last decade which inclJde some or all of these databases. In particularly, during the last 5 years a number c! systems have been implemented which attempt to provide supportive learning environments intended to facilitate leaf-Pin-by In this paper, we address a particular aspect of the problem of inferring a model from the pupil's behaviour on a set of tasks.

Report 80-28.pdf

Classics (Collection 2)

Portions of this report were reprinted with the permission of Mark Stefik. This report contains an overview oi the UNIT Package, a guide to the use of the Unit Editor (UE), a summary of the contents of the Bootstrap knowledge base, and a surnmnry of high-level access functions. It also contains implementation information, including a discussion of the underlying data structures, global variables and low-level functions. This reasearch was upported by the National Science Foundation under grant mcs78-02777 and by the Departmeant of National Defense of Canada, Research and Developrr -,nt Branch. Portions of Chapter Two were copies with permission from several reports by Mark Stefik.

Report 78-07.pdf

Classics (Collection 2)

COMPUTER SCIENCE DEPARTMENT School of Humanities and Sciences STANFORD UNIVERSITY DISTRIBUTED PROBLEM SOLVING: THE CONTRACT NET APPROACH Reid G. Smith and Randall Davis STAN-CS- 78_667 Heuristic Programming Project Memo 78-7 ABSTRACT We describe a problem solver based on a group of processor nodes which cooperate to solve problems. In a departure from earlier systems, we view task distribution as an interactive process, a discussion carried on between a node with a task to be executed and a group of nodes that may be able to execute the task. This leads to the use of a control formalism based on a contract metaphor, in which task distribution corresponds to contract negotiation. We also consider the kinds of knowledge that are used in such a problem solver, the way that the knowledge is indexed within an individual node, and distributed among the group of nodes. We suggest two primary methods of indexing the knowledge (referred to as "task-centered" and "knowledge-source centered"), and show how both methods can be useful.


Classics (Collection 2)

Overview and Bibliography of Distributed Databases. OVERVIEW AND BIBLIOGRAPHY OF DISTRIBUTED DATA BASES Hector Garcia-Molina 1 Department of Computer Science Stanford University Stanford, California, 94305 1 INTRODUCTION The goal of this paper is to provide an overview and a critical bibliography for the area of distributed data bases. Because of the recent - echnological advances in computer networks and communications, and because of the cost reduction of computer hardware, there has been a great interest in distributed data bases including some attempts at actual implementations. In this paper, we will first define what we mean by a distributed data base. Then we will give some of the reasons why people are so interested in this new field.

The Inference of Regular LISP Programs from Examples

Classics (Collection 2)

It is shown that the algorithm has robust performance for a wide variety of inputs and that it converges to a solution on the basis of minimum input information. IE DREAM of many a computer programmer has been a system that would accept a few randomly chosen examples of the desired behavior and that would immediately print out a general program for achieving the desired behavior in all possible situations. Thus the system user might type in the input-output pairs (6, 13), (2, 3), (7, 17), and (1, 2) and expect to have the system type out a program that reads an integer i (such as 6) and then prints the ith prime number (13 in this case). The system has the incredibly difficult task of discovering from this weak source of information 1) an algorithm for doing the desired computation, and 2) a correct implementation of the algorithm in some programming language. It is, in fact, doubtful that any system could ever exist which could create a class of large and general programs from such minimal information within a practical amount of time.

Sequencing and scheduling: Algorithms and complexity


In Graves, S. C., Zipkin, P. H., and Kan, A. H. G. R. (Eds.), Logistics of Production and Inventory: Handbooks in Operations Research and Management Science, Volume 4, pp. 445–522. North-Holland.