Gaussian processes for fast policy optimisation of POMDP-based dialogue managers

Modelling dialogue as a Partially Observable Markov Decision Process (POMDP) enables a dialogue policy robust to speech understanding errors to be learnt. However, a major challenge in POMDP policy learning is to maintain tractability, so the use of approximation is inevitable. We propose applying Gaussian Processes in Reinforcement learning of optimal POMDP dialogue policies, in order (1) to make the learning process faster and (2) to obtain an estimate of the uncertainty of the approximation. We first demonstrate the idea on a simple voice mail dialogue task and then apply this method to a real-world tourist information dialogue task. © 2010 Association for Computational Linguistics.

Safety implications of reactivity variations in fast thorium ADSRs

Nuclear power generation offers a reliable, low-impact and large-scale alternative to fossil fuels. However, concerns exist over the safety and sustainability of this method of power production, and it remains unpopular with some governments and pressure groups throughout the world. Fast thorium fuelled accelerator-driven sub-critical reactors (ADSRs) offer a possible route to providing further re-assurance regarding these concerns on account of their properties of enhanced safety through sub-critical operation combined with reduced actinide waste production from the thorium fuel source. The appropriate sub-critical margin at which these reactors should operate is the subject of continued debate. Commercial interests favour a small sub-critical margin in order to minimise the size of the accelerator needed for a given power output, whilst enhanced safety would be better satisfied through larger sub-critical margins to further minimise the possibility of a criticality excursion. Against this background, this paper examines some of the issues affecting reactor safety inherent within thorium fuel sources resulting from the essential Th90232→Th90233→Pa91233→U92233 breeding chain. Differences in the decay half-lives and fission and capture cross-sections of 233Pa and 233U can result in significant changes in the reactivity of the fuel following changes in the reactor power. Reactor operation is represented using a homogeneous lumped fast reactor model that can simulate the evolution of actinides and reactivity variations to first-order accuracy. The reactivity of the fuel is shown to increase significantly following a loss of power to the accelerator. Where the sub-critical operating margins are small this can result in a criticality excursion unless some form of additional intervention is made, for example through the insertion of control rods. © 2012 Elsevier Ltd. All rights reserved.

Spoken Dialog Challenge 2010: Comparison of live and control test results

The Spoken Dialog Challenge 2010 was an exercise to investigate how different spoken dialog systems perform on the same task. The existing Let's Go Pittsburgh Bus Information System was used as a task and four teams provided systems that were first tested in controlled conditions with speech researchers as users. The three most stable systems were then deployed to real callers. This paper presents the results of the live tests, and compares them with the control test results. Results show considerable variation both between systems and between the control and live tests. Interestingly, relatively high task completion for controlled tests did not always predict relatively high task completion for live tests. Moreover, even though the systems were quite different in their designs, we saw very similar correlations between word error rate and task completion for all the systems. The dialog data collected is available to the research community. © 2011 Association for Computational Linguistics.

Centrifugal modelling of the landslides triggering mechanism in layered fill slopes

Concerns over loosely compacted fill slopes stability in Hong Kong arouse in the past few decades, since the Sau Mau Ping disasters in 1972 and 1976. Research conducted on loose fill slopes in the past few years aimed to understand the failure mechanisms of a loosely compacted fill slope. Recently, layering effect has been hypothesised to be a possible condition in the fill slopes leading to a fast flowslide triggered by a rise of water table. Centrifuge experiments were conducted to investigate the layering effect on a model granular slope and hence to determine the triggering mechanisms of seepage induced slope failure. Test results showed that slope failure can be easily triggered in layered fill model slopes when seepage is restricted and localised pore water pressure is allowed to build up within the slope. © 2006 Taylor & Francis Group, London.

A validation of commercial flow solver performance on hybridised octree meshes

In recent years we have been developing a meshing system which is aimed at eliminating the bottleneck represented by building meshes for real-world, complex turbomachinery configurations. This system is based on a rapid octree meshing technology which is then made conformal to the bodies present. The objective of this paper is to demonstrate that this class of mesh is not only very fast to produce but also fit-for-purpose in the sense that simulations generated with third-party commercial flow solvers like Fluent have the same accuracy as those performed on more conventional meshes. A range of standard examples and test cases will be presented. Copyright © 2011 by ASME.

Occurrence of particle debris field during focused Ga ion beam milling of glassy carbon

To explore the machining characteristics of glassy carbon by focused ion beam (FIB), particles induced by FIB milling on glassy carbon have been studied in the current work. Nano-sized particles in the range of tens of nanometers up to 400 nm can often be found around the area subject to FIB milling. Two ion beam scanning modes - slow single scan and fast repetitive scan - have been tested. Fewer particles are found in single patterns milled in fast repetitive scan mode. For a group of test patterns milled in a sequence, it was found that a greater number of particles were deposited around sites machined early in the sequence. In situ EDX analysis of the particles showed that they were composed of C and Ga. The formation of particles is related to the debris generated at the surrounding areas, the low melting point of gallium used as FIB ion source and the high contact angle of gallium on glassy carbon induces de-wetting of Ga and the subsequent formation of Ga particles. Ultrasonic cleaning can remove over 98% of visible particles. The surface roughness (Ra) of FIB milled areas after cleaning is less than 2 nm. © 2010.

Stiffness of sands through a laboratory test database

Deformations of sandy soils around geotechnical structures generally involve strains in the range small (0·01%) to medium (0·5%). In this strain range the soil exhibits non-linear stress-strain behaviour, which should be incorporated in any deformation analysis. In order to capture the possible variability in the non-linear behaviour of various sands, a database was constructed including the secant shear modulus degradation curves of 454 tests from the literature. By obtaining a unique S-shaped curve of shear modulus degradation, a modified hyperbolic relationship was fitted. The three curve-fitting parameters are: an elastic threshold strain γe, up to which the elastic shear modulus is effectively constant at G0; a reference strain γr, defined as the shear strain at which the secant modulus has reduced to 0·5G0; and a curvature parameter a, which controls the rate of modulus reduction. The two characteristic strains γe and γr were found to vary with sand type (i.e. uniformity coefficient), soil state (i.e. void ratio, relative density) and mean effective stress. The new empirical expression for shear modulus reduction G/G0 is shown to make predictions that are accurate within a factor of 1·13 for one standard deviation of random error, as determined from 3860 data points. The initial elastic shear modulus, G0, should always be measured if possible, but a new empirical relation is shown to provide estimates within a factor of 1·6 for one standard deviation of random error, as determined from 379 tests. The new expressions for non-linear deformation are easy to apply in practice, and should be useful in the analysis of geotechnical structures under static loading.

Investigations of current limiting properties of the MgB2 wires subjected to pulse overcurrents in the benchtop tester

A laboratory scale desktop test system including a cryogenic system, an AC pulse generation system and a real time data acquisition program in LabView/DAQmx, has been developed to evaluate the quench properties of MgB 2 wires as an element in a superconducting fault current limiter under pulse overcurrents at 25K in self-field conditions. The MgB2 samples started from a superconducting state and demonstrated good current limiting properties characterized by a fast transition to the normal state during the first half of the cycle and a continuously limiting effect in the subsequent cycles without burnouts. The experimental and numerical simulation results on the quench behaviour indicate the feasibility of using MgB 2 for future superconducting fault current limiter (SFCL) applications. © IOP Publishing Ltd.

The Silent Period of Evidence Integration in Fast Decision Making

In a typical experiment on decision making, one out of two possible stimuli is displayed and observers decide which one was presented. Recently, Stanford and colleagues (2010) introduced a new variant of this classical one-stimulus presentation paradigm to investigate the speed of decision making. They found evidence for "perceptual decision making in less than 30 ms". Here, we extended this one-stimulus compelled-response paradigm to a two-stimulus compelled-response paradigm in which a vernier was followed immediately by a second vernier with opposite offset direction. The two verniers and their offsets fuse. Only one vernier is perceived. When observers are asked to indicate the offset direction of the fused vernier, the offset of the second vernier dominates perception. Even for long vernier durations, the second vernier dominates decisions indicating that decision making can take substantial time. In accordance with previous studies, we suggest that our results are best explained with a two-stage model of decision making where a leaky evidence integration stage precedes a race-to-threshold process. © 2013 Rüter et al.