Abstract. We investigate a committee-based approach for active learning of real-valued functions. This is a variance-only strategy for selection of informative training data. As such it is shown to suffer when the model class is misspecified since the learner's bias is high. Conversely, the strategy outperforms passive selection when the model class is very expressive since active minimization of the variance avoids overfitting.
A representation of the search space in optical pulse shaping problems employing an acousto-optic programmable dispersive filter (AOPDF) is presented for use in closed-loop learning experiments where the optimal spectral phase function to some control problem is determined by an iterative learning algorithm. The representation allows the algorithm to select a value for the optical chirp at each frequency control point such that only acoustic grating functions which preserve the spectrum of the shaped pulses are tested. The limits of this space with respect to the rate of applied optical chirp, optical bandwidth and acoustic power are examined and tested through diffraction efficiency studies performed using a commercial AOPDF. The main benefits of this representation are the elimination of undesirable frequency mixing effects, reduction of diffraction efficiency variation between arbitrary pulse shapes and faster convergence of the evolutionary algorithm.
An autonomous mobile robot requires an onboard controller that allows it to perform its tasks for long periods in isolation. One possibility is for the robot to adapt to its environment using some form of artificial intelligence. Evolutionary techniques such as genetic programming (GP) offer the possibility of automatically programming the controller based on the robot's experience of the world. Grammatical evolution (GE) is a recent evolutionary algorithm that has been successfully applied to various problems, particularly those for which GP has been successful. We present a method for applying GE to autonomous robot control and evaluate it in simulation for the Khepera robot
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.