20 20 20
Home     Organization    Submission    Dates    Venue    Program    Registration    Support   


The 14th International Conference on
Modeling Decisions for Artificial Intelligence
人にやさしい人工知能のための意思決定モデルの
構築に関する第 14 回国際会議 (MDAI 2017)
Kitakyushu, Japan October 18 - 20, 2017
http://www.mdai.cat/mdai2017
原稿締切:
USB DEADLINE: June 22nd, 2017
USB Submission open

INVITED TALKS




Prof. Ehud Lehrer
Department of Statistics and Operations Research in Tel-Aviv University, Israel.
Integration of non-additive probabilities: aggregation when information is incomplete

Abstract: Quite often decision makers have only partial information about the underlying uncertainty. This might happen, for instance, when information about the subject matter is obtained from different surveys/resources. We model such an information as a non-additive probability. Consider a decision maker who has to choose between two portfolios, or between two groups of engineers, based on incomplete information about the uncertainty of the market, or about the productivity of the groups.How would the decision maker evaluate the expected return from each portfolio or expected productivity from each group? We present different schemes of aggregation with respect to non-additive probabilities. These schemes might serve as decision tools in many fields, such as financial markets, production and more.


Prof. Michio Sugeno
Tokyo Institute of Technology, Japan.
A Step toward Nonlinear Statistics

Abstract: In this study we present a first step toward nonlinear statistics by applying Choquet calculus to probability theory. Throughout the study we take a constructive approach. For nonlinear statistics, we consider a distorted probability space on the nonnegative real line. A distorted probability measure (Edwards 1953) is derived from a conventional probability measure by the monotone transformation with a generator (usually called a distortion function), where we deal with two classes of parametric generators. First we explore some properties of Choquet integrals of continuous functions with respect to distorted probabilities. Then we calculate basic statistics such as the distorted mean and variance of a random variable for uniform, exponential and Gamma distributions.

In general we can consider a fuzzy measure space(Sugeno 1974)for nonlinear statistics.


Prof. Masaaki Nagahara
The University of Kitakyushu, Japan.
Sparsity methods for estimation and control

Abstract: Recently, sparsity has been playing a central role in signal processing, machine learning, and data science. Here we consider a problem of reconstructing (or learning) a signal (or a function) from observed data, which may be under-sampled and disturbed by noise. To address this problem, a method called sparse modeling, also known as compressed sensing, has become a hot topic. In this talk, I will give a brief introduction to sparse modeling for signal estimation, and its applications to control. In particular, I will give an introduction to "maximum hands-off control," which has the minimum support length among all feasible solutions for saving energy and reducing CO2 emissions in control systems.


Prof. Hiroshi SAKAMOTO
Kyushu Institute of Technology (Japan)
Stream Data Compression and Its Applications

Abstract: Social networking service and sensing device have become more and more popular in recent years and data flow never stop to increase. Examples are genome sequences of same species, version controlled documents, and source codes in repositories. Since such a data is usually highly-compressible, adopting data compression techniques is a suitable way to process it. In addition, in order to catch up the speed of data grow, there is a strong demand for stream data compression, that is, fully online and really scalable compression. In this talk, I would like to focus on lossless data compression and introduce several state-of-the-art technologies for stream data compression including their applications.


Prof. Guillermo Navarro-Arribas
Universitat Autònoma de Barcelona, Catalonia, Spain.
Provenance and privacy

Abstract: This talk presents an overview of current needs on data provenance in relationship with data privacy. It discusses state-of-the-art results in the area. In particular, the paper discusses properties and representation of data provenance, secure data provenance, and data provenance systems. Then, the talk will highlight the difficulties that we need to face in data provenance in relation to data privacy, and includes some lines of research that require further work.



 
MDAI 2017

University of Skövde

MDAI - Modeling Decisions

Vicenç Torra, Last modified: 13 : 26 September 02 2017.