More than 60 people attended the first version of SETIA, on April 23, 2010. Two tutorial-type talks took place on the first half, followed by 4 specialized talks, with the participation of speakers from Chile and Denmark. The following is the list of talks, including abstacts and links to their slides:

 

  1. “Una Introducción Intuitiva a la Teoría de la Información” (en español)
    Slides
    Dr. Milan S. Derpich
    Departamento de Electrónica,
    Universidad Técnica Federico Santa María, Chile 

    Abstract:
    En esta charla tutorial se redescubrirá, a partir de la
    noción humana y cotidiana de información, las expresiones y cantidades
    básicas utilizadas en la Teoría de la Información. Mediante un razonamiento
    simple e intuitivo, se derivarán y discutirán las expresiones para la
    entropía, la entropía condicional y la información mutua, y se dará un vistazo
    a la propiedad de la equipartición asintótica y a algunas de sus
    consecuencias. Finalmente se enunciarán y discutirán algunos de los resultados
    más importantes de esta teoría, tales como le teorema de capacidad de canales
    sin ruido, el de capacidad con canales con ruido, el teorema de la separación
    y otros resultados asociados al compromiso entre tasa de datos y distorsión.

 

  1. “An Introduction To Multiple-Description Coding — A Joint Source-
    Channel Coding Paradigm and its Potential Application to Digital TV”
    Slides
    Dr. Jan Ostergaard,
    Aalborg University, Denmark 

    Abstract:
    In 1948 Claude E. Shannon provided a coding theorem for joint source-channel
    coding; basically stating that if the source coding rate is strictly below the
    channel capacity, then reliable communication is possible. Moreover, in the
    direct part of the proof, he showed that, for stationary memoryless sources
    and channels, it is in fact possible to perform separate source and channel
    coding without any loss. Specifically, the source coding can be done without
    taking the channel into account and the channel coding can be done without any
    knowledge of the source. However, in order to achieve such source-channel
    separation, encoding schemes of high delay and complexity are generally
    required. Moreover, today’s communication networks have a heterogeneous
    infrastructure which deviates from earlier point-to-point scenarios. Thus, the
    separation theorem is not directly applicable to these kind of situations and
    it is often possible to improve the performance by exploiting joint source-
    channel coding techniques.

    In this talk, we provide an introduction to a joint source-channel coding
    technique known as multiple-description (MD) coding. In MD coding, a single
    source is encoded into several descriptions. All the descriptions are able to
    individually approximate the source to within prescribed fidelities.
    Furthermore, the descriptions are able to refine each other and thus jointly
    provide improvements. The potential applicability of MD coding to digital
    video broadcast is also discussed.

 

  1. “Non-Product Data-Dependent  Partitions for Mutual Information Estimation: Strong Consistency and Applications”
    Slides
    Dr. Jorge Silva
    Department of Electrical Engineering
    Facultad de Ciencias Físicas y Matemáticas
    Universidad de Chile 

    Abstract:
    The problem of mutual information (MI) estimation based on data-dependent
    partition is addressed in this work. A histogram-based construction
    is  proposed,  considering non-product data-dependent partitions, and
    sufficient conditions are stipulated  to guarantee a strongly consistent
    estimate for mutual information.
    On the applications of this result two emblematic families of density-free
    strongly consistent estimates are derived, one based on statistically
    equivalent blocks  (the Gessaman’s partition) and the other, on a
    tree-structured vector quantization scheme.
    Preliminary experimental  results demonstrate the superiority
    of these data-driven techniques in terms of
    a bias-variance analysis  when compared to
    conventional product histogram-based and kernel plug-in estimates.

 

  1. Title: “A Framework for Control System Design Subject to Average Data-Rate Constraints”
    Slides
    Dr. Eduardo I. Silva
    Departamento de Electrónica
    Universidad Técnica Federico Santa María, Chile 

    Abstract:
    In this talk we will study the performance of control systems subject
    to average data-rate limits.  By focusing on a class of source coding schemes
    built around entropy coded dithered quantizers, we will describe a novel
    framework to deal with such constraints in a tractable manner that combines
    ideas from both information and control theories.  The focus is on a situation
    where a noisy linear system has been designed assuming transparent feedback
    and, due to implementation constraints, a source coding scheme (with unity
    signal transfer function) has to be deployed in the feedback path.  The aim is
    to design such coding scheme so as to minimize the impact of quantization on
    the variance of a certain error signal (e.g., tracking error).  For this
    problem, a closed form upper bound on the best achievable performance for a
    given average data-rate constraint will be given.  We will also study the
    interplay between stability and average data-rates for the considered
    architecture. It will be shown that the proposed class of coding schemes can
    achieve mean square stability at average data-rates that are, at most, 1.254
    bits per sample away from the absolute minimum rate for stability established
    by Nair and Evans.  This rate penalty is compensated by the simplicity of the
    proposed approach.

 

  1. “Improved Upper Bounds to the Causal Quadratic Rate-Distortion Function for Gaussian Stationary Sources”
    Slides
    Dr. Milan S. Derpich
    Departamento de Electrónica,
    Universidad Técnica Federico Santa María, Chile 

    Abstract:
    The minimum data rate required to encode a random source with mean squared
    error (MSE) distortion D is given by its rate-distortion function (RDF),
    commonly denoted by R(D). The RDF for Gaussian stationary sources was fully
    characterized shortly after Claude Shannon introduced the concept. However, it
    is well known that achieving this RDF requires the use of non-causal filters,
    which in practice implies unbounded delays. In contrast, much less is known
    about the RDF for such sources under the additional constraint of causality or
    zero delay. Causal and zero-delay coders are attractive in applications such
    as voice and video coding, as well as in feedback systems.

    In this talk, we improve the existing achievable rate regions for causal and
    for zero-delay source coding of stationary Gaussian sources for mean squared
    error (MSE) distortion. First, we define the information-theoretic causal rate-
    distortion function (RDF), R_c^{it}(D). In order to analyze Rc^{it}(D), we
    introduce \overline{R_c}^{it}(D), the information theoretic causal RDF when
    the reconstruction error is jointly stationary with the source. We then derive
    four closed form upper bounds to the gap between R_c^{it} (D) and Shannon’s
    RDF, two of them strictly smaller than 0.5 bits/sample at all rates, and show
    that \overline{R_c}^{it}(D) can be realized by an AWGN channel surrounded by a
    unique set of causal pre-, post-, and feedback filters. A key result is showing
    that finding such filters constitutes a convex optimization problem and propose
    an iterative procedure to solve it. Finally, we build upon \overline{R_c}^{it}(D)
    to improve existing bounds on the optimal performance attainable by causal
    and zero-delay codes. This talk presents the results in a paper recently
    accepted for ISIT2010.

 

  1. “Noise-Shaped Predictive Coding for Multiple Descriptions of a Colored Gaussian Source”
    Slides
    Dr. Jan Ostergaard
    Aalborg University, Denmark 

    Abstract:
    In this talk we address the connection between the multiple-description (MD)
    problem and Delta-Sigma quantization. Specifically, we exploit the inherent
    redundancy  due to oversampling in Delta-Sigma quantization, and the simple
    linear-additive noise model resulting from dithered lattice quantization, in
    order to construct a symmetric  MD coding scheme. We show that the use of
    feedback by means of a noise shaping filter makes it possible to trade off
    central distortion for side distortion.  We then turn our attention to
    Gaussian sources with memory. Specifically, we consider stationary (colored)
    Gaussian sources and combine noise shaping and source prediction.  We first
    propose a new representation for the test channel that realizes the MD rate-
    distortion function of a Gaussian source, both in the white and in the colored
    source case.  We then show that this test channel can be materialized by
    embedding two source prediction loops, one for each description, within a
    common noise shaping loop. While the noise shaping loop controls the trade-off
    between the side and the central distortions, the role of prediction (like in
    differential pulse code modulation) is to extract the source innovations from
    the reconstruction at each of the side decoders, and thus reduce the coding
    rate. Finally, we show that this scheme achieves the MD rate-distortion
    function at all resolutions and all side-to-central distortion ratios, in the
    limit of high dimensional quantization.