Two experiments were conducted to explore the associations involved in human predictive learning. Experiment 1 found that post-training devaluation of one of the outcomes by instructing participants that one of the attackers was indestructible, led to a significant decrease in participants’ predictive responses to the cue that was initially followed by the devalued outcome, suggesting that cue-outcome associations play a major role in human simple predictive learning. In Experiment 2, immediately after receiving cue-outcome predictive judgment training, participants were instructed to destroy the attackers by using the same responses previously used to give predictive judgments, but they were not informed as to which response should be used on each attacker. During a test in which both attackers were present at the same time, when the cue was present, participants preferentially chose the instrumental response alternative that was previously used as a predictive judgment about the relationship between the present cue and its outcome. In the absence of cues, participants equally chose either response alternative. This transfer of control shows that participants also establish judgment-outcome associations during predictive training.

Today's industrial machines and manipulators have no capability to learn by experience. Performance and productivity could be greatly enhanced if a machine could modify its operation based on previous actions. This paper presents a learning control scheme that provides the ability for machines to utilize their past experiences. The objective is to have machines mimic the human learning process as closely as possible. A data base is formulated to provide the machine with experience. An optical infrared distance sensor is developed to inform the machine about objects in its working space. A learning control scheme is presented that utilizes the sensory information to enhance machine performance in the next trial. An adaptive scheme is proposed for the modification of learning gain matrices, and is implemented on an industrial robot. Experimental results verify the potentials of the proposed adaptive learning scheme, and illustrate how it can be used for improvement of different manufacturing processes.

The application of machine learning (ML) to solve practical problems is complex. Only recently, due to the increased promise of ML in solving real problems and the experienced difficulty of their use, has this issue started to attract attention. This difficulty arises from the complexity of learning problems and the large variety of available techniques. In order to understand this complexity and begin to overcome it, it is important to construct a characterization of learning situations. Building on previous work that dealt with the practical use of ML, a set of dimensions is developed, contrasted with another recent proposal, and illustrated with a project on the development of a decision-support system for marine propeller design. The general research opportunities that emerge from the development of the dimensions are discussed. Leading toward working systems, a simple model is presented for setting priorities in research and in selecting learning tasks within large projects. Central to the development of the concepts discussed in this paper is their use in future projects and the recording of their successes, limitations, and failures.