Showing papers on "Active learning (machine learning) published in 1985"

Pattern-recognizing stochastic learning automata

Abstract: A class of learning tasks is described that combines aspects of learning automation tasks and supervised learning pattern-classification tasks. These tasks are called associative reinforcement learning tasks. An algorithm is presented, called the associative reward-penalty, or AR-P algorithm for which a form of optimal performance is proved. This algorithm simultaneously generalizes a class of stochastic learning automata and a class of supervised learning pattern-classification methods related to the Robbins-Monro stochastic approximation procedure. The relevance of this hybrid algorithm is discussed with respect to the collective behaviour of learning automata and the behaviour of networks of pattern-classifying adaptive elements. Simulation results are presented that illustrate the associative reinforcement learning task and the performance of the AR-P algorithm as compared with that of several existing algorithms.

Analysis and design of an optimal learning control scheme for industrial robots: A discrete system approach

Abstract: Analysis and design of a discrete control system that can improve its performance in the course of operation is described in this paper. Such a system is called "discrete learning systems." A new discrete learning algorithm for controlling electro-mechanical systems such as industrial robots is proposed. The algorithm utilizes the error of state variables of the system, which includes positional and velocity error. A condition of algorithmic convergence is obtained. Advantages of a discrete system approach over an analog approach are discussed. Similarity between a learning controller and an state observer is also discussed. In most cases the learning control law is not, uniquely obtained from the condition of algorithmic convergence. We apply system optimization schemes to learning control and identify the learning control law in several ways. We propose better forms for the gain which is defined uniquely. We also show that the gain obtained is a generalized inverse solution to the optimal learning control problem. A discrete learning control scheme based on the proposed algorithm is effectively applied to control an industrial robot.

Verification-based learning: a generalisation strategy for inferring problem-reduction methods

Abstract: A major impediment to the development of high-performance knowledge-based systems arises from the prohibitive effort involved in equipping these systems with a sufficient set of problem-solving methods. Thus, one important research problem in Machine Learning has been the study of techniques for inferring problem-solving methods from examples. Although a number of techniques for learning problem-solving methods have been described in the literature, all of them assume a state-space model of problem-solving. In this paper we describe a new technique for learning problem-reduction methods, Verification-Based Learning (VBL), which extends the earlier techniques to the problem-reduction formulation of problem-solving. We illustrate the VBL technique with examples drawn from circuit design and symbolic integration.

Machine Learning of Visual Recognition Using Genetic Algorithms

Abstract: The system consists of three steps. At the very outset some pre-processing are applied on the input image. Secondly face features are extracted, which will be taken as the input of the Back-propagation Neural Network (BPN) and Genetic Algorithm (GA) in the third step and classification is carried out by using BPN and GA. The proposed approaches are tested on a number of face images. Experimental results demonstrate the higher degree performance of these algorithms.

Learning Object-Level and Meta-Level Knowledge in Expert Systems.

Abstract: : A high performance expert system can be built by exploiting machine learning techniques. A learning model has been designed and implemented that is capable of constructing a knowledge base, in the form of rules, from a case library and continuously updating it to accommodate new facts. This model is designed primarily for EMYCIN-like systems in which there is uncertainty about data as well as about the strength of inference and in which the rules chain together to infer complex hypotheses. These features greatly complicate the learning problem. In machine learning, two issues that cannot be overlooked practically are efficiency and noise. A subprogram, called CONDENSER, is designed to remove irrelevant features during learning and improve the efficiency. The noise can be handled by optimizing the result to achieve minimal prediction errors. Another subprogram has been developed to learn meta-level rules which guide the invocation of object-level rules and thus enhance the performance of the expert system using the object-level rules. Using the ideas developed in this work, an expert program called JAUNDICE has been built, which can diagnose the likely disease and mechanisms of a patient with jaundice. Experiments with JAUNDICE show the developed theory and method of learning are effective in a complex and noisy environment where data may be inconsistent, incomplete, and erroneous. (Author)

Interactive effect of information structure sequence of information and process learning time on rule learning using computer-based instruction

Abstract: The purpose of this study was to investigate three instructional design variables hypothesized to improve rule learning by use of information processing methods. These variables included: analysis and structure of information, response-sensitive sequencing of information, and monitoring of learning time. Using secondary education students learning internal punctuation rules, results from two experiments showed that (a) structuring information by a schematic analysis improved learning over a taxonomic analysis, (b) a response-sensitive sequence that first adapted instruction for generalization and second discrimination improved learning over either sequence separately, and (c) program monitoring of the display time interval improved learning over learner control. Findings are discussed in reference to an interactive nature of learning theory, instructional systems, and computer technology.

A note on learning curve parameters

Abstract: This note describes differences in the feasible range of values that the learning curve parameter may take when a unit learning approach is used as opposed to a cumulative-average learning curve. Clarification of this point is important because learning curve theory now is used in a variety of areas.

Learning object-level and meta-level knowledge in expert systems (machine learning)

Abstract: A high performance expert system can be built by exploiting machine learning techniques. A learning model has been designed and implemented that is capable of constructing a knowledge base, in the form of rules, from a case library and continuously updating it to accommodate new facts. This model is designed primarily for EMYCIN-like systems in which there is uncertainty about data as well as about the strength of inference and in which the rules chain together to infer complex hypotheses. These features greatly complicate the learning problem. In machine learning, two issues that cannot be overlooked practically are efficiency and noise. A subprogram, called "CONDENSER", is designed to remove irrelevant features during learning and improve the efficiency. The noise can be handled by optimizing the result to achieve minimal prediction errors. Another subprogram has been developed to learn meta-level rules which guide the invocation of object-level rules and thus enhance the performance of the expert system using the object-level rules. Using the ideas developed in this work, an expert program called JAUNDICE has been built, which can diagnose the likely disease and mechanisms of a patient with jaundice. Experiments with JAUNDICE show the developed theory and method of learning are effective in a complex and noisy environment where data may be inconsistent, incomplete, and erroneous.

A prototypical approach to machine learning

Abstract: This paper presents an overview of a research programme on machine learning which is based on the fundamental process of categorization. A structure of a computer model designed to achieve categorization is outlined and the knowledge representational forms and developmental learning associated with this approach are discussed.

Learning systems with control situations recognition

Abstract: The application of recognition techniques to the control of processes with incomplete apriori information is discussed. The proposed structure is found to contain two subsystems: a learning controller acting as a control situations recognition system and a learning predictor having the role of an output situations recognition system. Finally the off- and on-line learning controller implementations are presented together with some design considerations.

Optimal Destabilisation, Active Learning, and the Choice of Step Length in Policy Reform

Abstract: A theoretical model is presented in which the decision maker's choice of step length for reform is allowed to depend on the consequences which this is expected to have for the accuracy of information about the structure of the economy. Also, destabilisation by means of introducing variance into policies, is desirable under some circumstances since it speeds up the learning process.