Algorithmique et complexité des systèmes à compteurs

Réalisé en cotutelle avec l'École normale supérieure de Cachan – Université Paris-Saclay

Algorithmique et complexité des systèmes à compteurs

Réalisé en cotutelle avec l'École normale supérieure de Cachan – Université Paris-Saclay

i-vector based speaker recognition on short utterances

Robust speaker verification on short utterances remains a key consideration when deploying automatic speaker recognition, as many real world applications often have access to only limited duration speech data. This paper explores how the recent technologies focused around total variability modeling behave when training and testing utterance lengths are reduced. Results are presented which provide a comparison of Joint Factor Analysis (JFA) and i-vector based systems including various compensation techniques; Within-Class Covariance Normalization (WCCN), LDA, Scatter Difference Nuisance Attribute Projection (SDNAP) and Gaussian Probabilistic Linear Discriminant Analysis (GPLDA). Speaker verification performance for utterances with as little as 2 sec of data taken from the NIST Speaker Recognition Evaluations are presented to provide a clearer picture of the current performance characteristics of these techniques in short utterance conditions.

The role of the harmonic vector average in motion integration

The local speeds of object contours vary systematically with the cosine of the angle between the normal component of the local velocity and the global object motion direction. An array of Gabor elements whose speed changes with local spatial orientation in accordance with this pattern can appear to move as a single surface. The apparent direction of motion of plaids and Gabor arrays has variously been proposed to result from feature tracking, vector addition and vector averaging in addition to the geometrically correct global velocity as indicated by the intersection of constraints (IOC) solution. Here a new combination rule, the harmonic vector average (HVA), is introduced, as well as a new algorithm for computing the IOC solution. The vector sum can be discounted as an integration strategy as it increases with the number of elements. The vector average over local vectors that vary in direction always provides an underestimate of the true global speed. The HVA, however, provides the correct global speed and direction for an unbiased sample of local velocities with respect to the global motion direction, as is the case for a simple closed contour. The HVA over biased samples provides an aggregate velocity estimate that can still be combined through an IOC computation to give an accurate estimate of the global velocity, which is not true of the vector average. Psychophysical results for type II Gabor arrays show perceived direction and speed falls close to the IOC direction for Gabor arrays having a wide range of orientations but the IOC prediction fails as the mean orientation shifts away from the global motion direction and the orientation range narrows. In this case perceived velocity generally defaults to the HVA.

Low-resource language recognition using a fusion of phoneme posteriorgram counts, acoustic and glottal-based i-vectors

This paper presents a description of our system for the Albayzin 2012 LRE competition. One of the main characteristics of this evaluation was the reduced number of available files for training the system, especially for the empty condition where no training data set was provided but only a development set. In addition, the whole database was created from online videos and around one third of the training data was labeled as noisy files. Our primary system was the fusion of three different i-vector based systems: one acoustic system based on MFCCs, a phonotactic system using trigrams of phone-posteriorgram counts, and another acoustic system based on RPLPs that improved robustness against noise. A contrastive system that included new features based on the glottal source was also presented. Official and postevaluation results for all the conditions using the proposed metrics for the evaluation and the Cavg metric are presented in the paper.

Exam 1981

Exam questions and solutions in LaTex

Exam 1979

Exam questions and solutions in LaTex

MA167exam01pdf

Exam questions and solutions in PDF

Exam 1981 pdf

Exam questions and solutions in PDF

MA167exam01

Exam questions and solutions in LaTex. Diagrams for the questions are all together in the support.zip file, as .eps files

Forecasting models of retail rents

The authors model retail rents in the United Kingdom with use of vector-autoregressive and time-series models. Two retail rent series are used, compiled by LaSalle Investment Management and CB Hillier Parker, and the emphasis is on forecasting. The results suggest that the use of the vector-autoregression and time-series models in this paper can pick up important features of the data that are useful for forecasting purposes. The relative forecasting performance of the models appears to be subject to the length of the forecast time-horizon. The results also show that the variables which were appropriate for inclusion in the vector-autoregression systems differ between the two rent series, suggesting that the structure of optimal models for predicting retail rents could be specific to the rent index used. Ex ante forecasts from our time-series suggest that both LaSalle Investment Management and CB Hillier Parker real retail rents will exhibit an annual growth rate above their long-term mean.

Global analytic regularity for structures of co-rank one

We consider real analytic involutive structures V, of co-rank one, defined on a real analytic paracompact orientable manifold M. To each such structure we associate certain connected subsets of M which we call the level sets of V. We prove that analytic regularity propagates along them. With a further assumption on the level sets of V we characterize the global analytic hypoellipticity of a differential operator naturally associated to V. As an application we study a case of tube structures.

Ethics in a capitalistic society

The problem of semantics is inherent in any discussion of ethics. The general term "ethics" is itself commonly confused. In addition, systems of ethics must be built upon assumptions, and assumptions are necessarily subject to lengthy debate. These two problems are encountered in my investigation of the ethical practices of the modern business community and to remedy the situation I have taken two steps: the first being an attempt to clarify the meaning of terms used therein;-and the second being a clear description of the assumptions utilized to further my analysis. To satisfy those who would disagree with these assumptions, I have attempted to outline the consequences of differing premises. The first assumption in my discussion is that the capitalistic economy is powered by the motivation supplied by man's self-interest. We are conditioned to basing our courses of action upon an orientation toward gratifying this self-interest. Careers are chosen by blending aptitude, interest, and remuneration. of course, some people are less materially inclined than others, but the average member of our capitalistic society is concerned with the physical rewards derived from his employment. Status and happiness are all-important considerations in pursuing a chosen course of action, yet all too often they are measured in physical terms. The normal self-interest natural to mankind is heightened in capitalism, due to the emphasis placed upon material compensation. Our thinking becomes mechanistic as life devolves into a complex game played by the rules. We are accustomed to performing meaningless or unpleasant duties to fulfill our gratifications. Thought, consequently, interferes with the completion of our everyday routines. We learn quickly not to be outspoken, as the outspoken one threatens the security of his fellow man. The majority of the people are quite willing to accept others views on morality, and indeed this is the sensible thing to do as one does not risk his own neck. The unfortunate consequence of this situation has been the substitution of the legal and jural for the moral and ethical. Our actions are guided by legal considerations and nowhere has this been more evident than in the business community. The large legal departments of modern corporations devote full time to inspecting the legality of corporate actions. The business community has become preoccupied with the law, yet this is necessarily so. Complex, modern, capitalistic society demands an elaborate framework of rules and regulations. Without this framework it would be impossible to have an orderly economy, to say nothing of protecting the best interests of the people. However, the inherent complexities, contradictions, and sometimes unfair aspects of our legal system can tempt men to take things into their own hands. From time to time cases arise where men have broken laws while acting in good faith, and other cases where men have been extremely unethical without being illegal. Examples such as these foster the growth of cynicism, and generally create an antagonistic attitude toward the law on the part of business. My second assumption is that the public, on the whole, has adopted an apathetic attitude toward business morality. when faced with an ethical problem, far too many people choose to cynically assume that, if I don't do it someone else will. "The danger of such an assumption lies in that it eliminates many of the inhibitions that normally would preclude unethical action. The preventative factor in contemplating an unethical act not only lies in it going against the "right course of action", but also in that it would display the actor as one of the few, immoral practitioners. However, if the contemplator feels that many other people follow the same course of action, he would not feel himself to be so conspicuous. These two assumptions underly my entire discussion of modern business ethics., and in my judgment are the two most important causal factors in unethical acts perpetrated by the business community. The future elimination of these factors seems improbable, if not futile, yet there is no reason to consider things worse than they ever have been before. The heightened public interest in business morality undoubtedly lies in part in the fact that examples of corporate malpractice are of such magnitude in scope, and hence more newsworthy.

Language recognition using phonotactic-based shifted delta coefficients and multiple phone recognizers

A new language recognition technique based on the application of the philosophy of the Shifted Delta Coefficients (SDC) to phone log-likelihood ratio features (PLLR) is described. The new methodology allows the incorporation of long-span phonetic information at a frame-by-frame level while dealing with the temporal length of each phone unit. The proposed features are used to train an i-vector based system and tested on the Albayzin LRE 2012 dataset. The results show a relative improvement of 33.3% in Cavg in comparison with different state-of-the-art acoustic i-vector based systems. On the other hand, the integration of parallel phone ASR systems where each one is used to generate multiple PLLR coefficients which are stacked together and then projected into a reduced dimension are also presented. Finally, the paper shows how the incorporation of state information from the phone ASR contributes to provide additional improvements and how the fusion with the other acoustic and phonotactic systems provides an important improvement of 25.8% over the system presented during the competition.