2,507
个编辑
更改
跳到导航
跳到搜索
第44行：
第44行： − 部分信息分解的具体计算方法目前仍没有普遍的共识，学界有非常多的计算定义被提出\cite. + − 下表总结了目前常见方法在不同案例下的计算结果差异： − + − + − + − + + + + + + + + + − + + + + + + + + + − + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
→部分信息分解
[[文件:Lattice_of_2.png|替代=|无框]]
[[文件:Lattice_of_2.png|替代=|无框]]
部分信息分解的具体计算方法目前仍没有普遍的共识，学界有非常多的计算定义被提出。下表总结了目前常见方法在不同案例下的计算结果差异：
*
*
* \begin{itemize} <nowiki> \item $I_{\cap}^{\text{*}}$, the redundancy measure proposed by Artemy Kolchinsky \cite{kolchinsky2022novel}.</nowiki> <nowiki> \item $I_{\cap}^{\text{WB}}$, the redundancy measure first proposed by Williams and Beer \cite{williams2010nonnegative}.</nowiki> \item $I_{\cap}^{\text{MMI}}$, the ``minimum mutual information<nowiki>''</nowiki> \cite{barrett2015exploration}. \item $I_{\cap}^{\wedge}$, proposed by Griffith et al. \cite{griffith2014intersection}. <nowiki> \item $I_{\cap}^{\text{GH}}$, proposed by Griffith and Ho \cite{griffith2015quantifying}.</nowiki> <nowiki> \item $I_{\cap}^{\text{Ince}}$, proposed by Ince \cite{ince2017measuring}.</nowiki> <nowiki> \item $I_{\cap}^{\text{FL}}$, proposed by Finn and Lizier \cite{finn2018pointwise}.</nowiki> <nowiki> \item $I_{\cap}^{\text{BROJA}}$, proposed by Bertschinger et al. \cite{bertschinger2014quantifying}.</nowiki> <nowiki> \item $I_{\cap}^{\text{Harder}}$, proposed by Harder et al. \cite{harder2013bivariate}.</nowiki> <nowiki> \item $I_{\cap}^{\text{dep}}$, proposed by James et al. \cite{james2018unique}.</nowiki> \end{itemize}
* \begin{itemize} <nowiki> \item $I_{\cap}^{\text{*}}$, the redundancy measure proposed by Artemy Kolchinsky \cite{kolchinsky2022novel}.</nowiki> <nowiki> \item $I_{\cap}^{\text{WB}}$, the redundancy measure first proposed by Williams and Beer \cite{williams2010nonnegative}.</nowiki> \item $I_{\cap}^{\text{MMI}}$, the ``minimum mutual information<nowiki>''</nowiki> \cite{barrett2015exploration}. \item $I_{\cap}^{\wedge}$, proposed by Griffith et al. \cite{griffith2014intersection}. <nowiki> \item $I_{\cap}^{\text{GH}}$, proposed by Griffith and Ho \cite{griffith2015quantifying}.</nowiki> <nowiki> \item $I_{\cap}^{\text{Ince}}$, proposed by Ince \cite{ince2017measuring}.</nowiki> <nowiki> \item $I_{\cap}^{\text{FL}}$, proposed by Finn and Lizier \cite{finn2018pointwise}.</nowiki> <nowiki> \item $I_{\cap}^{\text{BROJA}}$, proposed by Bertschinger et al. \cite{bertschinger2014quantifying}.</nowiki> <nowiki> \item $I_{\cap}^{\text{Harder}}$, proposed by Harder et al. \cite{harder2013bivariate}.</nowiki> <nowiki> \item $I_{\cap}^{\text{dep}}$, proposed by James et al. \cite{james2018unique}.</nowiki> \end{itemize}
{| class="wikitable"
{| class="wikitable"
|+
|+
!11
!<math>目标变量</math>
!12
!<math>I_{\cap}^* </math><ref>Artemy Kolchinsky. A novel approach to the partial information decomposition. Entropy, 24(3):403, 2022.</ref>
!13
!<math>\{\{1\}\}</math><ref>Paul L Williams and Randall D Beer. Nonnegative decomposition of multivariate information. arXiv preprint arXiv:1004.2515, 2010.</ref>
!
!<math>\{\{1\}\}</math>
!<math>\{\{1\}\}</math>
!<math>\{\{1\}\}</math>
!<math>\{\{1\}\}</math>
!<math>\{\{1\}\}</math>
!<math>\{\{1\}\}</math>
!<math>\{\{1\}\}</math>
!<math>\{\{1\}\}</math>
!<math>\{\{1\}\}</math>
|-
|-
|21
|<math>Y = X_1 \text{ AND } X_2</math>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|-
|-
|22
|<math>Y = X_1 + X_2</math>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|-
|-
|<math>Y = X_1</math>
|
|
|
|
|
|
|
|
|
|
|
|-
|<math>Y = (X_1, X_2)</math>
|
|
|
|
|
|
|
|
|
|
|