Humans appear to be able to learn new concepts without needing to be programmed explicitly in any conventional sense. In this paper we regard learning as the phenomenon of knowledge acquisition in the absence of explicit programming. We give a precise methodology for studying this phenomenon from a computational viewpoint. It consists of choosing an appropriate information gathering mechanism, the learning protocol, and exploring the class of concepts that can be learnt using it in a reasonable (polynomial) number of steps. We find that inherent algorithmic complexity appears to set serious limits to the range of concepts that can be so learnt. The methodology and results suggest concrete principles for designing realistic learning systems.

This paper presents the design and an initial performance evaluation of the query optimizer generator designed for the EXODUS extensible database system. Algebraic transformation rules are translated into an executable query optimizer, which transforms query trees and selects methods for executing operations according to cost functions associated with the methods. The search strategy avoids exhaustive search and it modifies itself to take advantage of past experience. Computational results show that an optimizer generated for a relational system produces access plans almost as good as those produced by exhaustive search, with the search time cut to a small fraction.

The long-term goal of our field is the creation and understanding of intelligence. Productive research in AI, both practical and theoretical, benefits from a notion of intelligence that is precise enough to allow the cumulative development of robust systems and general results. This paper outlines a gradual evolution in our formal conception of intelligence that brings it closer to our informal conception and simultaneously reduces the gap between theory and practice. 1 Artificial Intelligence AI is a field in which the ultimate goal has often been somewhat ill-defined and subject to dispute. Some researchers aim to emulate human cognition, others aim at the creation of

Learning control involves modifying a controller's behavior to improve its performance as measured by some predefined index of performance (IP). If control actions that improve performance as measured by the IP are known, supervised learning methods, or methods for learning from examples, can be used to train the controller. But when such control actions are not known a priori, appropriate control behavior has to be inferred from observations of the IP. One can distinguish between two classes of methods for training controllers under such circumstances. Indirect methods involve constructing a model of the problem's IP and using the model to obtain training information for the controller. On the other hand, direct, or model-free,...

Many graphics tasks, such as the manipulation of graphical objects, and the construction of userinterface widgets, can be facilitated by geometric constraints. However, the difficulty of specifying constraints by traditional methods forms a barrier to their widespread use. In order to make constraints easier to declare, we have developed a method of specifying constraints implicitly, through multiple examples. Snapshots are taken of an initial scene configuration, and one or more additional snapshots are taken after the scene has been edited into other valid configurations. The constraints that are satisfied in all the snapshots are then applied to the scene objects. We discuss an efficient algorithm for inferring constraints from multiple snapshots. The algorithm has been incorporated into the Chimera editor, and several examples of its use are discussed. 1 Introduction Geometric constraints are used extensively in computer graphics in the specification of relationships between graph...

Given a drawing of straight lines on the plane, we wish to decide whether it is the projec-tion of the visible part of a set of opaque polyhedra. This is the fundamental algorithmic problem that underlies much of the research in computer vision. Although there are extensive literature and reports on empirically successful algorithms for this problem and its many extensions, there has been no definite result concerning its complexity. In this paper we show that, rather surprisingly, this problem is NP-complete, and therefore there is probably no polynomial-time algorithm for solving it. This is true even in the relatively simple case of trihedral scenes (no four planes share a point) without shadows and cracks. Despite this negative result, we present positive results for the important special case of orthohedral scenes (ah planes are normal to one of the three axes). 6 1988 Academic Press, Inc. 1.

Given a sequence of events (or ob]ects), each 'characterized by a set of attributes, the problem considered is to discover a rule characterizing the sequence and able to predict a plausible sequence continuation. The rule, called a sequence-generating rule, is nondeterministic in the sense that it does not necessarily tell exactly which etent must appear next in the sequence, but rather, defines a set of plausible next eents. The basic assumption of the methodology presented here is that the next etent depends solely on the attributes of the previous eents in the sequence. These attributes are either initially given or can be den'td from the initial ones through a chain of inferences. Three basic rule models are employed to guide the search for a sequence.generating rule: decomposition, periodic, and disjunctive normal form (DNF). The search process involves simultaneously transforming the initial sequences to derived sequences and instantiating models to find the best match between the instantiated model and the derived sequence. A program, called SPARC/E, is described that implements most of the methodology a.v applied to discosring sequence generating rules in the card game Eleusis. This game, which models the process of scientiftc discovery, is used as a sottrce of examples for illustrating the performance of SPARC/E.

Abstract not found

Graphical editing, like many applications facilitated by computers, often involves repetitive tasks. To reduce repetition, programmers can write procedures to automate these tasks, however most users do not know how to program,

This paper presents a system for learning object-specific recognition strategies from training images and libraries of image understanding routines. The motivation for this work is that thirty years of computer vision research has produced hundreds of algorithms for visual subtasks ranging from edge detection to pose determination, but very few complete vision systems. The Schema Learning System (SLS) addresses this problem by casting object recognition as an control problem: for every object to be recognized, it learns a sequence of algorithms that will find it quickly and robustly. More formally, SLS learns control policies under supervision. For every task, a user specifies the target representation (e.g. 2D image position or 3D world position), and provides a set of training images and the locations of the target objects. SLS then applies a three-step process of search, learning from examples and graph optimization to produce a recognition graph that expresses a control policy for ...