Abstract. A new class of evolutionary computation processes is presented, called Learnable Evolution Model or LEM. In contrast to Darwinian-type evolution that relies on mutation, recombination, and selection operators, LEM employs machine learning to generate new populations. Specifically, in Machine Learning mode, a learning system seeks reasons why certain individuals in a population (or a collection of past populations) are superior to others in performing a designated class of tasks. These reasons, expressed as inductive hypotheses, are used to generate new populations. A remarkable property of LEM is that it is capable of quantum leaps (“insight jumps”) of the fitness function, unlike Darwinian-type evolution that typically proceeds through numerous slight improvements. In our early experimental studies, LEM significantly outperformed evolutionary computation methods used in the experiments, sometimes achieving speed-ups of two or more orders of magnitude in terms of the number of evolutionary steps. LEM has a potential for a wide range of applications, in particular, in such domains as complex optimization or search problems, engineering design, drug design, evolvable hardware, software engineering, economics, data mining, and automatic programming.

An enormous proliferation of databases in almost every area of human endeavor has created a great demand for new, powerful tools for turning data into useful, task-oriented knowledge. In efforts to satisfy this need, researchers have been exploring ideas and methods developed in machine learning, pattern recognition, statistical data analysis, data visualization, neural nets, etc. These efforts have led to the emergence of a new research area, frequently called data mining and knowledge discovery. The first part of this chapter is a compendium of ideas on the applicability of symbolic machine learning methods to this area. The second part describes a multistrategy methodology for conceptual data exploration, by which we mean the derivation of high-level concepts and descriptions from data through symbolic reasoning involving both data and background knowledge. The methodology, which has been implemented in the INLEN system, combines machine learning, database and knowledge-based techn...

The purpose of a document is to facilitate the transfer of information from its author to its readers. It is the author’s job to design the document so that the information it contains can be interpreted accurately and efficiently. To do this, the author can make use of a set of stylistic tools. In this paper we introduce the concept of document functionality, which attempts to describe the roles of documents and their components in the process of transferring information. A functional description of a document provides insight into the type of the document, into its intended uses, and into strategies for automatic document interpretation and retrieval. To demonstrate these ideas, we define a taxonomy of

Introduction In concept learning and data mining, a typical objective is to determine concept descriptions or patterns that will classify future data points as correctly as possible. If one can assume that the data contain no noise, then it is desirable that descriptions are complete and consistent with regard to all the data, i.e., they characterize all data points in a given class (positive examples) and no data points outside the class (negative examples). In real-world applications, however, data may be noisy, that is, they may contain various kinds of errors, such as errors of measurement, classification or transmission, and/or inconsistencies. In such situations, searching for consistent and complete descriptions ceases to be desirable. In the presence of noise, an increase in completeness (an increase of generality of a description) tends to cause a decrease in consistency and vice versa; therefore, the best strategy is to seek a description that represents the trade-off betwe

This report provides a description and a user's guide for AQ19, a program for machine learning and pattern discovery. AQ19 works in two modes: Theory Formation and Pattern Discovery. In Theory Formation mode, given examples of two or more concepts, AQ19 hypothesizes general descriptions of these concepts optimized according to a modifiable criterion of hypothesis preference. In Pattern Discovery mode, given data with indicated input and output variables, AQ19 determines strong patterns in the relationship between the input and output variables...

LEM3 is the newest implementation of Learnable Evolution Model (LEM), a non-Darwinian evolutionary computation methodology that employs machine learning to guide evolutionary processes. Due to a deep integration of different modes of operation and the use of the advanced machine learning system AQ21, the LEM3 system is a highly efficient and effective implementation of the methodology. LEM3 supports different attribute types for describing individuals in the population, such as nominal, rank, structured, interval and ratio, which makes it applicable to a wide range of practical problems. It also implements very efficient methods for switching between different modes of operation and operators controlling the generation of new individuals. This paper describes the underlying LEM3 algorithm, results from LEM3 testing on selected benchmark function optimization problems (with the number of variables varying from 10 to 1000), and its comparison with EA, a conventional, Darwinian-type evolutionary computation program. In every experiment, without exception, LEM3 outperformed EA in terms of the evolution length (the number of fitness evaluations needed to achieved a desired solution), sometimes very significantly. It also outperformed the previous LEM2 implementation.

AQ21 is a multitask machine learning and data mining system for attributional rule learning, rule testing, and application to a wide range of classification problems. One of its distinctive features is that it strives to perform natural induction, that is, it seeks inductive hypotheses that are not only accurate but also easy to understand and interpret. Although the system provides the user with a large number of control parameters, they all can be omitted if one wants to run it in default mode. The parameters control the mode of learning (theory formation vs. pattern discovery), the degree of generality of learned rules, and a range of preference criteria for selecting candidate rules that tailor the learning process to the given problem. AQ21 includes event and episode classification programs (ATEST and EPIC) that give the user control over the testing and application of the learned rulesets to the task at hand.

This paper presents work in progress on an approach to the problem of recognizing blasting caps in x-ray images. An analysis of functional properties of blasting caps was used to design the representation space, which combines intensity and shape features. Recognition proceeds in two phases. The first phase is a bottomup process in which low intensity blobs are used as attention-catching devices to generate object hypotheses. The second phase is a top-down process in which object hypotheses are confirmed or rejected by fitting a local model to ribbons surrounding the low intensity blob. The local model is acquired using inductive learning. Flexible matching routines are used during recognition that provide a measure of confidence for the identification. Experimental results demonstrate the ability to learn the relationship between image characteristics and object functionality. 1 Introduction This paper presents work in progress on an approach to the problem of recognizing blasting ca...

In this paper, we describe a method for learning shape descriptions of objects in x-ray images. The descriptions are induced from shape examples using the AQ15c inductive learning system. The method has been experimentally compared to k-nearest neighbor, a statistical pattern recognition technique, the C4.5 decision tree learning program, and a multilayer feed-forward neural network. Experimental results demonstrate strong advantages of the AQ methodology over the other methods. Specifically, the method has higher predictive accuracy and faster learning and recognition rates. AQ's representation language, VL 1 , was better suited for this problem, which can be seen by examining the empirical results and the learned rules. The method was applied to the problem of detecting blasting caps in x-ray images of luggage. An intelligent system performing this detection task can be used to assist airport security personnel with luggage screening. 1. INTRODUCTION Despite many efforts, the proble...

This chapter concerns problems of learning patterns in images and image sequences, and using them for interpreting new images. The chapter concentrates on three problem areas: (i) semantic interpretation of color images of outdoor scenes, (ii) detection of blasting caps in x-ray images of luggage, and (iii) recognizing actions in video image sequences. It discusses the image formation processes in these problem areas, and the choices of representation spaces used in our approaches to solving these problems. The results presented indicate the advantages of applying machine learning to vision. 10.1 INTRODUCTION The underlying motivation of this research is that vision systems need learning capabilities for handling problems for which algorithmic solutions are unknown or difficult to obtain. Learning capabilities can also make vision systems more easily adaptable to different vision problems, and more flexible and robust in handling variable perceptual conditions [MRA94]. Much of the cur...