Intelligent Behavior Utilizing Light-weight Diagnosis

Intelligence in its decisions is a trait that we have grown to expect from a cyber-physical system. In particular that it makes the right choices at runtime, i.e., those that allow it fulfill its tasks, even in case of faults or unexpected interactions with its environment. Analyzing how to continuously achieve the currently desired (and possibly continuously changing) goals and adapting its behavior to reach these goals is undoubtedly a serious challenge. This becomes even more challenging if the atomic actions a system can implement become unreliable due to faulty components or some exogenous event out of its control. In this paper, we propose a solution for the presented challenge. In particular, we show how to adopt a light-weight diagnosis concept to cope with such situations. The approach is based on rules coupled with means for rule selection that are based on previous information regarding the success or failure of rule executions. We furthermore present a Java-based framework of the light-weight diagnosis concept, and discuss the results obtained from an experimental evaluation considering several application scenarios. At the end, we present a qualitative comparison with other related approaches that should help the reader decide which approach works best for them.

IHAC: Incorporating Heuristics for Efficient Rule Generation & Rule Selection in Associative Classification

Associations rule mining along with classification rule mining are both significant techniques of mining of knowledge in the area of knowledge discovery in massive databases stored in different geographic locations of the world. Based on such combination of these two, class association rules for mining or associative classification methods have been generated, which, in far too many cases, showed higher prediction accuracy than platitudinous conventional classifiers. Motivated by the study, in this paper, we proposed a new approach, namely IHAC (Incorporating Heuristics for efficient rule generation & rule selection in Associative Classification). First, it utilises the database to decrease the search space and then explicitly explores the potent class association rules from the optimised database. This also blends rule generation and classifier building to speed up the overall classifier construction cycle. Experimental findings showed that IHAC performs better than any further associative classification methods.

Clustering and Dispatching Rule Selection Framework for Batch Scheduling

In this study, a batch scheduling with job grouping and batch sequencing is considered. A clustering algorithm and dispatching rule selection model is developed to minimize total tardiness. The model and algorithm are based on the constrained k-means algorithm and neural network. We also develop a method to generate a training dataset from historical data to train the neural network. We use numerical examples to demonstrate that the proposed algorithm and model efficiently and effectively solve batch scheduling problems.

Can We Agree on What Robots Should be Allowed to Do? An Exercise in Rule Selection for Ethical Care Robots

Future Care Robots (CRs) should be able to balance a patient’s, often conflicting, rights without ongoing supervision. Many of the trade-offs faced by such a robot will require a degree of moral judgment. Some progress has been made on methods to guarantee robots comply with a predefined set of ethical rules. In contrast, methods for selecting these rules are lacking. Approaches departing from existing philosophical frameworks, often do not result in implementable robotic control rules. Machine learning approaches are sensitive to biases in the training data and suffer from opacity. Here, we propose an alternative, empirical, survey-based approach to rule selection. We suggest this approach has several advantages, including transparency and legitimacy. The major challenge for this approach, however, is that a workable solution, or social compromise, has to be found: it must be possible to obtain a consistent and agreed-upon set of rules to govern robotic behavior. In this article, we present an exercise in rule selection for a hypothetical CR to assess the feasibility of our approach. We assume the role of robot developers using a survey to evaluate which robot behavior potential users deem appropriate in a practically relevant setting, i.e., patient non-compliance. We evaluate whether it is possible to find such behaviors through a consensus. Assessing a set of potential robot behaviors, we surveyed the acceptability of robot actions that potentially violate a patient’s autonomy or privacy. Our data support the empirical approach as a promising and cost-effective way to query ethical intuitions, allowing us to select behavior for the hypothetical CR.

Extrapolation accuracy underestimates rule learning: Evidence from the function-learning paradigm

Understanding the development of non-linear processes such as economic or populationgrowth is an important prerequisite for informed decisions in those areas. In the function-learningparadigm, people’s understanding of the function rule that underlies the to-be predicted process istypically measured by means of extrapolation accuracy. Here we argue, however, that even thoughaccurate extrapolation necessitates rule-learning, the reverse does not necessarily hold: Inaccurateextrapolation does not exclude rule-learning. Experiment 1 shows that more than one third of participants who would be classified as “exemplar-based learners” based on their extrapolation accuracy were able to identify the correct function shape and slope in a rule-selection paradigm, demonstrating accurate understanding of the function rule. Experiment 2 shows that higher proportions of rule learning than rule-application in the function learning paradigm is not due to (i) higher a priori probabilities to guess the correct rule in the rule-selection paradigm; nor is it due to (ii) a lack of simultaneous access to all function values in the function-learning paradigm. We conclude that rule application is not tantamount to rule-learning, and that assessing rule-learning via extrapolation accuracy underestimates the proportion of rule learners in function-learning experiments.