2016 Fiscal Year Final Research Report
A new development in Bayesian model averaging from the viewpoints of thermodynamics
| Project/Area Number |
26330044
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Statistical science
|
|---|
| Research Institution | Kyushu University |
|---|
Principal Investigator |
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
Dunn Peter University of the Sunshine Coast, Associate Professor
|
|---|
| Co-Investigator(Renkei-kenkyūsha) |
YANAGIMOTO Takemi 統計数理研究所, 名誉教授 (40000195)
|
|---|
| Project Period (FY) |
2014-04-01 – 2017-03-31
|
| Keywords | 統計科学 / Bayesモデル平均 / Bayes予測問題 / 熱力学 |
|---|
| Outline of Final Research Achievements |
We formulate a prediction problem, which is an estimation of a probability density, as a risk minimization problem. Here we adopt the alpha-divergence as a loss function for evaluating the goodness of a prediction. The alpha-divergence is a generalization of the Kullback-Leibler divergence. Through this minimization problem, we reveal a duality between the likelihood maximization and the Shannon entropy maximization, which are two of the most important principles in statistical science.
|
| Free Research Field |
統計科学
|
