The local adaptive processing of signals and images in a transform domain within a sliding window suggests certain advantages in some signal and image de-noising applications due to incorporating an available a priori information about the signals and noises. However, an optimum transform size is also data dependent and generally is not known in advance. Performing the de-noising with the varyi...

In this lecture we will talk about the PAC model. The PAC learning model is one of the important and famous learning model. PAC stands for Probably Approximately Correct, our goal is to learn a hypothesis from a hypothesis class such that in high con dence we will have a small error rate (approximately correct). We start the lecture with an intuitive example to explain the idea behind the PAC m...

We propose a new approach to developing a tractable affective dialogue model for general probabilistic frame-based dialogue systems. The dialogue model, based on the Partially Observable Markov Decision Process (POMDP) and the Dynamic Decision Network (DDN) techniques, is composed of two main parts, the slot level dialogue manager and the global dialogue manager. Our implemented dialogue manage...

This paper explores the use of a statistical technique known as density estimation to potentially improve the results of text categorization systems which label documents by computing similarities between documents and categories. In addition to potentially improving a system's overall accuracy, density estimation converts similarity scores to probabilities. These probabilities provide con denc...

We provide a refoundation of the symmetric growth equilibrium characterizing the research sector of vertical R&D-driven growth models. We argue that the usual assumptions made in this class of models leave the agents indi¤erent as to where targeting research: hence, the problem of the allocation of R&D investment across sectors is indeterminate. By introducing an “" contamination of con…dence” ...

In this paper we consider the problem of constructing exact conndence intervals for the common mean of several normal populations with unknown and possibly unequal variances. Several procedures based on pivots and P-values are discussed and compared.

In this paper, we study partial identi cation and inference for a general class of functionals of the joint distribution of potential outcomes of a binary treatment under the strong ignorability assumption or the selection on observables assumption commonly used in evaluating average treatment e ects. Members of this class of functionals include the correlation coe cient between the potential o...

This paper proposes an utterance veri cation system for hidden Markov model (HMM) based automatic speech recognition systems. A veri cation objective function, based on a multi-layer-perceptron (MLP), is adopted which combines con dence measures from both the recognition and veri cation models. Discriminative minimum veri cation error training is applied for optimizing the parameters of the MLP...

Direct measures of expectations, derived from survey data, are used in a Vector Autoregressive model of actual and expected output in eight industries in the UK manufacturing sector. No evidence is found with which to reject rationality in the expectations series when measurement error is appropriately taken into account. The VAR analysis illustrates the importance of intersectoral interactions...

A stochastic program SP with solution value z∗ can be approximately solved by sampling n realizations of the program’s stochastic parameters, and by solving the resulting “approximating problem” for (x∗ n ; z ∗ n ). We show that, in expectation, z ∗ n is a lower bound on z∗ and that this bound monotonically improves as n increases. The rst result is used to construct con dence intervals on the ...