Competitive training in hidden Markov models

The use of hidden Markov models is placed in a connectionist framework, and an alternative approach to improving their ability to discriminate between classes is described. Using a network style of training, a measure of discrimination based on the a posteriori probability of state occupation is proposed, and the theory for its optimization using error back-propagation and gradient ascent is presented. The method is shown to be numerically well behaved, and results are presented which demonstrate that when using a simple threshold test on the probability of state occupation, the proposed optimization scheme leads to improved recognition performance.

Recurrent input transformations for Hidden Markov models

This paper presents a new architecture which integrates recurrent input transformations (RIT) and continuous density HMMs. The basic HMM structure is extended to accommodate recurrent neural networks which transform the input observations before they enter the Gaussian output distributions associated with the states of the HMM. During training the parameters of both HMM and RIT are simultaneously optimized according to the Maximum Mutual Information (MMI) criterion. Results are presented for the E-set recognition task which demonstrate the ability of recurrent input transformations to exploit longer term correlations in the speech signal and to give improved discrimination.

APPROXIMATE GAUSS-MARKOV REALISATION OF MULTIVARIATE STOCHASTIC PROCESSES.

Given a spectral density matrix or, equivalently, a real autocovariance sequence, the author seeks to determine a finite-dimensional linear time-invariant system which, when driven by white noise, will produce an output whose spectral density is approximately PHI ( omega ), and an approximate spectral factor of PHI ( omega ). The author employs the Anderson-Faurre theory in his analysis.

Blind equalization and sequence estimation using Markov chain Monte Carlo methods

This paper discusses the problem of restoring a digital input signal that has been degraded by an unknown FIR filter in noise, using the Gibbs sampler. A method for drawing a random sample of a sequence of bits is presented; this is shown to have faster convergence than a scheme by Chen and Li, which draws bits independently. ©1998 IEEE.

On the relationship between hidden Markov Models and convex functional transforms for simulating scanning probe Microscopy

Models for simulating Scanning Probe Microscopy (SPM) may serve as a reference point for validating experimental data and practice. Generally, simulations use a microscopic model of the sample-probe interaction based on a first-principles approach, or a geometric model of macroscopic distortions due to the probe geometry. Examples of the latter include use of neural networks, the Legendre Transform, and dilation/erosion transforms from mathematical morphology. Dilation and the Legendre Transform fall within a general family of functional transforms, which distort a function by imposing a convex solution.In earlier work, the authors proposed a generalized approach to modeling SPM using a hidden Markov model, wherein both the sample-probe interaction and probe geometry may be taken into account. We present a discussion of the hidden Markov model and its relationship to these convex functional transforms for simulating and restoring SPM images.©2009 SPIE.

A markov chain Monte Carlo method for estimating population mixing using Y-chromosome markers: Mixing of the Han people in China

We present a new approach for estimating mixing between populations based on non-recombining markers, specifically Y-chromosome microsatellites. A Markov chain Monte Carlo (MCMC) Bayesian statistical approach is used to calculate the posterior probability

Policies on poverty and the fisheries sector in Uganda: the current policy trends

There are concerns, at least among the proponents of development, on how to link policy development processes in Uganda and the associated transformation of the poor to high standards of living. In fact some questions have been posed as to whether it's the absence of poverty-targeted policies that a good proportion of individuals or communities are still poor. In the fisheries sector where most of the fish dependent communities live, poverty indications are still prevalent although arguments have been put that current reforms in the sector have transformed the lives of the fish dependent communities. The 1999/2000 household survey report indicates that the poverty levels reduced to 35% of Uganda's total population from 44% in 1997. The question that arose, which still arises anyway, was to define who is actually poor. When measuring poverty one is ultimately interested in the 'standards of living' of individuals especially those, whose standards of living are inadequate. The basic element of measuring this inadequacy/adequacy, at least in Uganda, is to use the household income or consumption per adult equivalent. Studies have demonstrated that household consumption expenditure is a good approximation of household income1. Therefore, for purpose of this report, we define poor households to mean based on that that one adopted by the Ministry of Finance to mean "households whose expenditure per adult equivalent falls below the poverty line 3 ". Many government documents report that the poverty line is one dollar a day. Therefore someone is below the poverty line if he or she lives on less than one dollar a day. In this paper, we analyse the evolution of poverty-driven policies that have been put in place by government and how these policies are shifting or are likely to shift the lives of fish dependent communities. We argue that combinations of poverty-policies are being translated into increased incomes and welfare of most individuals in the fisheries sector. The reasons for this shift, we argue, is as a result of a combination of factors all supported by non other that poverty-led government policies.

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.

Partially Observable Markov Decision Processes with continuous observations for dialogue management

This work shows how a dialogue model can be represented as a Partially Observable Markov Decision Process (POMDP) with observations composed of a discrete and continuous component. The continuous component enables the model to directly incorporate a confidence score for automated planning. Using a testbed simulated dialogue management problem, we show how recent optimization techniques are able to find a policy for this continuous POMDP which outperforms a traditional MDP approach. Further, we present a method for automatically improving handcrafted dialogue managers by incorporating POMDP belief state monitoring, including confidence score information. Experiments on the testbed system show significant improvements for several example handcrafted dialogue managers across a range of operating conditions.