Results


Planning and Meta-Planning (MOLGEN: Part 2)

Classics (Collection 2)

The selection of what to do next is often the hardest part of resource-limited problem solving. In planning problems, there are typically many goals to be achieved in some order. The goals interact with each other in ways which depend both on the order in which they are achieved and on the particular operators which are used to achieve them. A planning program needs to keep its options open because decisions about one part of a plan are likely to have consequences for another part. This paper describes an approach to planning which integrates and extends two strategies termed the least-commitment and the heuristic strategies.


loo

Classics (Collection 2)

Remember the Library! they need your suggestions to service your needs Artificial Intelligence Review 10: 441-475,1996. In this paper we discuss a hybrid approach combining Case-Based Reasoning (CBR) . Remember the Library! they need your suggestions to service your needs nuwer Waltham; Nigel Shadbolt, University of Nottingham; Aaron Sloman, University of Birmingham; Austin Tate, Edinburgh University Ray Turner, University of Essex; David Warren, University of Bristol; Yorick Wilks, New Mexico State University Artificial Intelligence Review offers state-of-the-art survey and tutorial papers and provides a forum for review and commentary on Artificial Intelligence foundations, applications and current research. Artificial Intelligence Review includes refereed commentary offering multiple perspectives on significant developments in Artificial Intelligence, Cognitive Science and related disciplines. It aims to illustrate the variety of styles and points of view which characterises ...


10 An Experiment on Inductive Learning in Chess End Games

Classics (Collection 2)

Further progress in the application of computers to many practical fields seems to depend heavily on the success in implementing learning and inductive processes within machines. For example, to develop a consultation system for medical or plant disease diagnosis, prognosis and decision making in general, it is very desirable, perhaps even necessary, to be able to'teach' the system through examples of correct and/or incorrect decisions, rather than by precisely describing the decision process in its full generality and then transforming this description into a computer program. A similar situation exists in computer chess. The development of computer programs playing at the master level (especially the end games) seems to be a formidable task if the programs are not eventually able to learn and improve on their decision making rules through the specific examples of games, rather than by being explicitly told all the rules. Due to easy access to human knowledge about chess and the relative simplicity of testing the results, chess is one of the most attractive testing domains for inductive inference programs.


BOXES: AN EXPERIMENT IN ADAPTIVE CONTROL

Classics (Collection 2)

BOXES is the name of a computer program. This is what the chess player does when he lumps together large numbers of positions as being'similar' to each other, by neglecting the strategically irrelevant features in which they differ. The resultant small game can be said to be a'model' of the large game. To give a brutally extreme example, consider a specification of chess positions so incomplete as to map from the viewpoint of White the approximately 1050 positions of the large game on to the seven shown in Figure 1. Even this simple classification may have a role in the learning of chess.


Experiments in Machine Learning and Protein Folding

Classics (Collection 2)

The aim of these experiments is to test the use of machine learning as a tool for forming theories from data. A machine-learning program (PRorms) was developed to form rules for predicting protein secondary structure from primary structure--an important unsolved problem in molecular biology. Structured background knowledge is used to transform the search space and control generalization. Six rules were found that are humanly comprehensible and provide a chemically meaningful description of the important factors in formation of secondary structure. Variations of the rules were found with different accuracies. These were found to help highlight the important features of the rule. Rules were also found which used threshold logic to match sequences of primary structure. These rules were found to be suitable for predicting turn secondary structure. Perhaps the most promising and yet most difficult application of machine learning is in the area of scientific discovery: 'the most technically ...


Error Tolerant Learning Systems

Classics (Collection 2)

We consider the task of a robot learning in a reactive environment by performing experiments. A reactive environment is one where changes occur in response to actions. Actors other than the learner may be present in the world. The robot performs experiments by modifying the environment and observing the outcome. These observations lead to a collection of concepts which constitute a theory of the behaviour of the environment, also called a world model.


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.


11 Incremental Learning of Concept Descriptions: A Method and Experimental Results

Classics (Collection 2)

A system for learning concept descriptions incrementally is described and illustrated by a series of experiments in the domains of insect classification, chess endgames and plant disease diagnosis. The system employs a full-memory learning method that incrementally improves hypotheses, but does not forget facts. The method is used to form both characteristic descriptions, which describe a concept in detail, and discriminant descriptions, which specify only properties needed to distinguish a given concept from a given set of other concepts. Experimental results show the advantages of inducing and maintaining only characteristic descriptions during learning and creating discriminant descriptions from them when a classification decision is necessary. Research in the area of concept learning from examples has been concerned mainly with methods for single step, or non-incremental, learning.


Semi-autonomousacquisition of pattern-based knowledge

Classics (Collection 2)

The task of acquiring and organizing the knowledge on which to base an expert system is difficult. These themes are explored in the context of attempts to construct high-performance programs relevant to the chess endgame king-rook versus king-knight. Most existing expert systems are based on knowledge obtained from a human expert. With reference to the family of geological expert systems being built at SRI, Gaschnig writes: Model development is a cooperative enterprise involving an exploration geologist who is an authority on the type of deposit being modelled, and a computer scientist who understands the operation of the PROSPECTOR system [1]. Feigenbaum, one of the pioneers of expert systems work and head of probably the world's largest group building such systems, puts it as follows: (The knowledge engineer) works intensively with an expert to acquire domain-specific knowledge and organise it for use by a program [2].