Abstract :
Summary form only given, as follows. English abstract provided: In practical scenes of data analyses, we did not always assume an ideal objective result is derived based on objective dataset. Rather, we often dare to include in the analysis results subjective biases based on the analysers?? own experiences and/or arbitrary view points, and thus we cannot necessarily be confident on those subjective results. The theory of Bayesian probability provides a reasonable way to achieve a result involving both objective observations and subjective biases, and the Bayesian modeling techniques derive such a moderate model that efficiently uses small amount of objective data. In my talk, I will explain the basic concepts which are needed to understand the framework of Bayesian modeling, and introduce several computational techniques such as sampling methods, variational approximation, and expectation propagation.
