Effects of the user model on simulation-based learning of dialogue strategies

Over the past decade, a variety of user models have been proposed for user simulation-based reinforcement-learning of dialogue strategies. However, the strategies learned with these models are rarely evaluated in actual user trials and it remains unclear how the choice of user model affects the quality of the learned strategy. In particular, the degree to which strategies learned with a user model generalise to real user populations has not be investigated. This paper presents a series of experiments that qualitatively and quantitatively examine the effect of the user model on the learned strategy. Our results show that the performance and characteristics of the strategy are in fact highly dependent on the user model. Furthermore, a policy trained with a poor user model may appear to perform well when tested with the same model, but fail when tested with a more sophisticated user model. This raises significant doubts about the current practice of learning and evaluating strategies with the same user model. The paper further investigates a new technique for testing and comparing strategies directly on real human-machine dialogues, thereby avoiding any evaluation bias introduced by the user model. © 2005 IEEE.

Model based learning of sigma points in unscented Kalman filtering

The unscented Kalman filter (UKF) is a widely used method in control and time series applications. The UKF suffers from arbitrary parameters necessary for a step known as sigma point placement, causing it to perform poorly in nonlinear problems. We show how to treat sigma point placement in a UKF as a learning problem in a model based view. We demonstrate that learning to place the sigma points correctly from data can make sigma point collapse much less likely. Learning can result in a significant increase in predictive performance over default settings of the parameters in the UKF and other filters designed to avoid the problems of the UKF, such as the GP-ADF. At the same time, we maintain a lower computational complexity than the other methods. We call our method UKF-L. ©2010 IEEE.

Model based learning of sigma points in unscented Kalman filtering

The unscented Kalman filter (UKF) is a widely used method in control and time series applications. The UKF suffers from arbitrary parameters necessary for sigma point placement, potentially causing it to perform poorly in nonlinear problems. We show how to treat sigma point placement in a UKF as a learning problem in a model based view. We demonstrate that learning to place the sigma points correctly from data can make sigma point collapse much less likely. Learning can result in a significant increase in predictive performance over default settings of the parameters in the UKF and other filters designed to avoid the problems of the UKF, such as the GP-ADF. At the same time, we maintain a lower computational complexity than the other methods. We call our method UKF-L. © 2011 Elsevier B.V.

Altruistic learning.

The origin of altruism remains one of the most enduring puzzles of human behaviour. Indeed, true altruism is often thought either not to exist, or to arise merely as a miscalculation of otherwise selfish behaviour. In this paper, we argue that altruism emerges directly from the way in which distinct human decision-making systems learn about rewards. Using insights provided by neurobiological accounts of human decision-making, we suggest that reinforcement learning in game-theoretic social interactions (habitisation over either individuals or games) and observational learning (either imitative of inference based) lead to altruistic behaviour. This arises not only as a result of computational efficiency in the face of processing complexity, but as a direct consequence of optimal inference in the face of uncertainty. Critically, we argue that the fact that evolutionary pressure acts not over the object of learning ('what' is learned), but over the learning systems themselves ('how' things are learned), enables the evolution of altruism despite the direct threat posed by free-riders.

A modular planar robotic manipulandum with end-point torque control.

Robotic manipulanda are extensively used in investigation of the motor control of human arm movements. They permit the application of translational forces to the arm based on its state and can be used to probe issues ranging from mechanisms of neural control to biomechanics. However, most current designs are optimized for studying either motor learning or stiffness. Even fewer include end-point torque control which is important for the simulation of objects and the study of tool use. Here we describe a modular, general purpose, two-dimensional planar manipulandum (vBOT) primarily optimized for dynamic learning paradigms. It employs a carbon fibre arm arranged as a parallelogram which is driven by motors via timing pulleys. The design minimizes the intrinsic dynamics of the manipulandum without active compensation. A novel variant of the design (WristBOT) can apply torques at the handle using an add-on cable drive mechanism. In a second variant (StiffBOT) a more rigid arm can be substituted and zero backlash belts can be used, making the StiffBOT more suitable for the study of stiffness. The three variants can be used with custom built display rigs, mounting, and air tables. We investigated the performance of the vBOT and its variants in terms of effective end-point mass, viscosity and stiffness. Finally we present an object manipulation task using the WristBOT. This demonstrates that subjects can perceive the orientation of the principal axis of an object based on haptic feedback arising from its rotational dynamics.

Perceptual learning, roving and the unsupervised bias

Perceptual learning improves perception through training. Perceptual learning improves with most stimulus types but fails when . certain stimulus types are mixed during training (roving). This result is surprising because classical supervised and unsupervised neural network models can cope easily with roving conditions. What makes humans so inferior compared to these models? As experimental and conceptual work has shown, human perceptual learning is neither supervised nor unsupervised but reward-based learning. Reward-based learning suffers from the so-called unsupervised bias, i.e., to prevent synaptic " drift" , the . average reward has to be exactly estimated. However, this is impossible when two or more stimulus types with different rewards are presented during training (and the reward is estimated by a running average). For this reason, we propose no learning occurs in roving conditions. However, roving hinders perceptual learning only for combinations of similar stimulus types but not for dissimilar ones. In this latter case, we propose that a critic can estimate the reward for each stimulus type separately. One implication of our analysis is that the critic cannot be located in the visual system. © 2011 Elsevier Ltd.

Identification of Materials from Construction Site Images Using Content Based Image Retrieval Techniques

The capability to automatically identify shapes, objects and materials from the image content through direct and indirect methodologies has enabled the development of several civil engineering related applications that assist in the design, construction and maintenance of construction projects. Examples include surface cracks detection, assessment of fire-damaged mortar, fatigue evaluation of asphalt mixes, aggregate shape measurements, velocimentry, vehicles detection, pore size distribution in geotextiles, damage detection and others. This capability is a product of the technological breakthroughs in the area of Image and Video Processing that has allowed for the development of a large number of digital imaging applications in all industries ranging from the well established medical diagnostic tools (magnetic resonance imaging, spectroscopy and nuclear medical imaging) to image searching mechanisms (image matching, content based image retrieval). Content based image retrieval techniques can also assist in the automated recognition of materials in construction site images and thus enable the development of reliable methods for image classification and retrieval. The amount of original imaging information produced yearly in the construction industry during the last decade has experienced a tremendous growth. Digital cameras and image databases are gradually replacing traditional photography while owners demand complete site photograph logs and engineers store thousands of images for each project to use in a number of construction management tasks. However, construction companies tend to store images without following any standardized indexing protocols, thus making the manual searching and retrieval a tedious and time-consuming effort. Alternatively, material and object identification techniques can be used for the development of automated, content based, construction site image retrieval methodology. These methods can utilize automatic material or object based indexing to remove the user from the time-consuming and tedious manual classification process. In this paper, a novel material identification methodology is presented. This method utilizes content based image retrieval concepts to match known material samples with material clusters within the image content. The results demonstrate the suitability of this methodology for construction site image retrieval purposes and reveal the capability of existing image processing technologies to accurately identify a wealth of materials from construction site images.