Extending noise robust structured support vector machines to larger vocabulary tasks

This paper describes a structured SVM framework suitable for noise-robust medium/large vocabulary speech recognition. Several theoretical and practical extensions to previous work on small vocabulary tasks are detailed. The joint feature space based on word models is extended to allow context-dependent triphone models to be used. By interpreting the structured SVM as a large margin log-linear model, illustrates that there is an implicit assumption that the prior of the discriminative parameter is a zero mean Gaussian. However, depending on the definition of likelihood feature space, a non-zero prior may be more appropriate. A general Gaussian prior is incorporated into the large margin training criterion in a form that allows the cutting plan algorithm to be directly applied. To further speed up the training process, 1-slack algorithm, caching competing hypothesis and parallelization strategies are also proposed. The performance of structured SVMs is evaluated on noise corrupted medium vocabulary speech recognition task: AURORA 4. © 2011 IEEE.

Rao-Blackwellized particle smoothers for mixed linear/nonlinear state-space models

We consider the smoothing problem for a class of conditionally linear Gaussian state-space (CLGSS) models, referred to as mixed linear/nonlinear models. In contrast to the better studied hierarchical CLGSS models, these allow for an intricate cross dependence between the linear and the nonlinear parts of the state vector. We derive a Rao-Blackwellized particle smoother (RBPS) for this model class by exploiting its tractable substructure. The smoother is of the forward filtering/backward simulation type. A key feature of the proposed method is that, unlike existing RBPS for this model class, the linear part of the state vector is marginalized out in both the forward direction and in the backward direction. © 2013 IEEE.

Transient natural ventilation of a space with localised heating

The natural ventilation of a well-mixed, pre-heated room with a point source of heating, and openings at the base and roof is investigated. The transient draining associated with the room being warmer than the exterior combined with the convective ow produced by the point source of heat leads to a fascinating series of transient ow regimes as the system evolves to the two-layer steady-state regime described by Linden, Lane-Ser_ and Smeed [1]. As the room begins to ventilate, a turbulent plume rises from the point source of heat to the ceiling, and typically forms a deepening layer of hot air. However, with a weak heat source, then at some point the ascending plume will intrude beneath the layer of original uid. Otherwise, the ascending plume always reaches the top of the room as the system evolves to a steady state. We develop a simpli_ed model of the transient evolution and test this with some new laboratory experiments. We conclude with a discussion of the implications of our results for real buildings.

Free-Space Interconnect for Optical Switch and Analogue Neural Network

In this paper, the architecture of a vector-matrix multiplier (MVM) is simulated. The optical design can be made compact by the use of GRIN lenses for the optical fan-in. The intended application area was in storage area networks (SANs) but the concept can be applied to a neural network. © 2011 Allerton Press, Inc.

Sequential decision making in general state space models

Many problems in control and signal processing can be formulated as sequential decision problems for general state space models. However, except for some simple models one cannot obtain analytical solutions and has to resort to approximation. In this thesis, we have investigated problems where Sequential Monte Carlo (SMC) methods can be combined with a gradient based search to provide solutions to online optimisation problems. We summarise the main contributions of the thesis as follows. Chapter 4 focuses on solving the sensor scheduling problem when cast as a controlled Hidden Markov Model. We consider the case in which the state, observation and action spaces are continuous. This general case is important as it is the natural framework for many applications. In sensor scheduling, our aim is to minimise the variance of the estimation error of the hidden state with respect to the action sequence. We present a novel SMC method that uses a stochastic gradient algorithm to find optimal actions. This is in contrast to existing works in the literature that only solve approximations to the original problem. In Chapter 5 we presented how an SMC can be used to solve a risk sensitive control problem. We adopt the use of the Feynman-Kac representation of a controlled Markov chain flow and exploit the properties of the logarithmic Lyapunov exponent, which lead to a policy gradient solution for the parameterised problem. The resulting SMC algorithm follows a similar structure with the Recursive Maximum Likelihood(RML) algorithm for online parameter estimation. In Chapters 6, 7 and 8, dynamic Graphical models were combined with with state space models for the purpose of online decentralised inference. We have concentrated more on the distributed parameter estimation problem using two Maximum Likelihood techniques, namely Recursive Maximum Likelihood (RML) and Expectation Maximization (EM). The resulting algorithms can be interpreted as an extension of the Belief Propagation (BP) algorithm to compute likelihood gradients. In order to design an SMC algorithm, in Chapter 8 uses a nonparametric approximations for Belief Propagation. The algorithms were successfully applied to solve the sensor localisation problem for sensor networks of small and medium size.

Sequential Monte Carlo computation of the score and observed information matrix in state-space models with application to parameter estimation

Sequential Monte Carlo (SMC) methods are popular computational tools for Bayesian inference in non-linear non-Gaussian state-space models. For this class of models, we propose SMC algorithms to compute the score vector and observed information matrix recursively in time. We propose two different SMC implementations, one with computational complexity $\mathcal{O}(N)$ and the other with complexity $\mathcal{O}(N^{2})$ where $N$ is the number of importance sampling draws. Although cheaper, the performance of the $\mathcal{O}(N)$ method degrades quickly in time as it inherently relies on the SMC approximation of a sequence of probability distributions whose dimension is increasing linearly with time. In particular, even under strong \textit{mixing} assumptions, the variance of the estimates computed with the $\mathcal{O}(N)$ method increases at least quadratically in time. The $\mathcal{O}(N^{2})$ is a non-standard SMC implementation that does not suffer from this rapid degrade. We then show how both methods can be used to perform batch and recursive parameter estimation.

The modified Kanerva model for automatic speech recognition

A parallel processing network derived from Kanerva's associative memory theory Kanerva 1984 is shown to be able to train rapidly on connected speech data and recognize further speech data with a label error rate of 0·68%. This modified Kanerva model can be trained substantially faster than other networks with comparable pattern discrimination properties. Kanerva presented his theory of a self-propagating search in 1984, and showed theoretically that large-scale versions of his model would have powerful pattern matching properties. This paper describes how the design for the modified Kanerva model is derived from Kanerva's original theory. Several designs are tested to discover which form may be implemented fastest while still maintaining versatile recognition performance. A method is developed to deal with the time varying nature of the speech signal by recognizing static patterns together with a fixed quantity of contextual information. In order to recognize speech features in different contexts it is necessary for a network to be able to model disjoint pattern classes. This type of modelling cannot be performed by a single layer of links. Network research was once held back by the inability of single-layer networks to solve this sort of problem, and the lack of a training algorithm for multi-layer networks. Rumelhart, Hinton & Williams 1985 provided one solution by demonstrating the "back propagation" training algorithm for multi-layer networks. A second alternative is used in the modified Kanerva model. A non-linear fixed transformation maps the pattern space into a space of higher dimensionality in which the speech features are linearly separable. A single-layer network may then be used to perform the recognition. The advantage of this solution over the other using multi-layer networks lies in the greater power and speed of the single-layer network training algorithm. © 1989.

Unsteady ejectors: The effect of driver jet mark-space ratio

mark Unsteady ejectors can be driven by a wide range of driver jets. These vary from pulse detonation engines, which typically have a long gap between each slug of fluid exiting the detonation tube (mark-space ratios in the range 0.1-0.2) to the exit of a pulsejet where the mean mass flow rate leads to a much shorter gap between slugs (mark-space ratios in the range 2-3). The aim of this paper is to investigate the effect of mark-space ratio on the thrust augmentation of an unsteady ejector. Experimental testing was undertaken using a driver jet with a sinusoidal exit velocity profile. The mean value, amplitude and frequency of the velocity profile could be changed allowing the length to diameter ratio of the fluid slugs L/D and the mark-space ratio (the ratio of slug length to the spacing between slugs) L/S to be varied. The setup allowed L/S of the jet to vary from 0.8 to 2.3, while the L/D ratio of the slugs could take any values between 3.5 and 7.5. This paper shows that as the mark-space ratio of the driver jet is increased the thrust augmentation drops. Across the range of mark-space ratios tested, there is shown to be a drop in thrust augmentation of 0.1. The physical cause of this reduction in thrust augmentation is shown to be a decrease in the percentage time over which the ejector entrains ambient fluid. This is the direct result ofthe space between consecutive slugs in the driver jet decreasing. The one dimensional model reported in Heffer et al. [1] is extended to include the effect of varying L/S and is shown to accurately capture the experimentally measured behavior ofthe ejector. Copyright © 2010 by the American Institute of Aeronautics and Astronautics, Inc.

Model-based approaches to handling additive noise in reverberant environments

Model-based approaches to handle additive and convolutional noise have been extensively investigated and used. However, the application of these schemes to handling reverberant noise has received less attention. This paper examines the extension of two standard additive/convolutional noise approaches to handling reverberant noise. The first is an extension of vector Taylor series (VTS) compensation, reverberant VTS, where a mismatch function including reverberant noise is used. The second scheme modifies constrained MLLR to allow a wide-span of frames to be taken into account and projected into the required dimensionality. To allow additive noise to be handled, both these schemes are combined with standard VTS. The approaches are evaluated and compared on two tasks, MC-WSJ-AV, and a reverberant simulated version of AURORA-4. © 2011 IEEE.

Measurements of triggering and transient growth in a model lean-premixed gas turbine combustor

Instability triggering and transient growth of thermoacoustic oscillations were experimentally investigated in combination with linear/nonlinear flame transfer function (FTF) methodology in a model lean-premixed gas turbine combustor operated with CH 4 and air at atmospheric pressure. A fully premixed flame with 10kW thermal power and an equivalence ratio of 0.60 was chosen for detailed characterization of the nonlinear transient behaviors. Flame transfer functions were experimentally determined by simultaneous measurements of inlet velocity fluctuations and heat release rate oscillations using a constant temperature anemometer and OH */CH * chemiluminescence emissions, respectively. The phase-resolved variation of the local flame structure at a limit cycle was measured by planar laser-induced fluorescence of OH. Simultaneous measurements of inlet velocity, OH */CH * emission, and acoustic pressure were performed to investigate the temporal evolution of the system from a stable to a limit cycle operation. This measurement allows us to describe an unsteady instability triggering event in terms of several distinct stages: (i) initiation of a small perturbation, (ii) exponential amplification, (iii) saturation, (iv) nonlinear evolution of the perturbations towards a new unstable periodic state, (v) quasi-steady low-amplitude periodic oscillation, and (vi) fully-developed high-amplitude limit cycle oscillation. Phase-plane portraits of instantaneous inlet velocity and heat release rate clearly show the presence of two different attractors. Depending on its initial position in phase space at infinitesimally small amplitude, the system evolves towards either a high-amplitude oscillatory state or a low-amplitude oscillatory state. This transient phenomenon was analyzed using frequency- and amplitude-dependent damping mechanisms, and compared to subcritical and supercritical bifurcation theories. The results presented in this paper experimentally demonstrate the hypothesis proposed by Preetham et al. based on analytical and computational solutions of the nonlinear G-equation [J. Propul. Power 24 (2008) 1390-1402]. Good quantitative agreement was obtained between measurements and predictions in terms of the conditions for the onset of triggering and the amplitude of triggered combustion instabilities. © 2011 The Combustion Institute.