Corrigendum: Modeling Emotions Associated With Novelty at Variable Uncertainty Levels: A Bayesian Approach

A Corrigendum on
Modeling Emotions Associated With Novelty at Variable Uncertainty Levels: A Bayesian Approach

by Yanagisawa, H., Kawamata, O., and Ueda, K. (2019). Front. Comput. Neurosci. 13:2. doi: 10.3389/fncom.2019.00002

In the original article, there were errors in Equations (5) and (12) and the text.

In Equation (5) there was a typographical error in the numerator and we wrote

$\begin{array}{l}{\eta }_{post}=\frac{s_p\eta +s_l\bar{x}}{s_p+s_l}\end{array}$

when it should be

$\begin{array}{l}{\eta }_{post}=\frac{s_p\bar{x}+s_l\eta }{s_p+s_l}\end{array}$

In Equation (12) it was written that uncertainty is always less than zero:

$\begin{array}{l}\frac{\partial \beta }{\partial s_p}=\frac{1}{2}\left\{-\frac{s_p}{s_l(s_p+s_l)}-\frac{s_l}{{(s_p+s_l)}^2}\right\}\le 0\end{array}$

This has been corrected to show that uncertainty s_p is always more than zero:

$\begin{array}{l}\frac{\partial \beta }{\partial s_p}=\frac{s_p}{2{(s_p+s_l)}^2}>0\end{array}$

Similarly, there was a miscalculation in the text preceding Equation (14). We wrote “We derived ${\delta }^2+({\alpha }_1-{\alpha }_2)/({\beta }_1+{\beta }_2)=0$ under β₁ ≠ β₂. Therefore, (α₁ − α₂)/(β₁ + β₂) < 0 it is the condition.” We actually derived “${\delta }^2({\alpha }_1-{\alpha }_2)+({\beta }_1-{\beta }_2)=0$ under β₁ ≠ β₂” and the condition is “(α₁ − α₂)(β₁ − β₂) < 0.”

A correction has been made to the Model of Emotional Dimensions Elicited by A Novel Event section, subsection Interaction Effect of Uncertainty and Prediction Errors on Information Gain, paragraph 5.

“A condition where the two functions have an intersection is ${\alpha }_1{\delta }^2+{\beta }_1$ = ${\alpha }_2{\delta }^2+{\beta }_2$. We derived ${\delta }^2({\alpha }_1-{\alpha }_2)+({\beta }_1-{\beta }_2)=0$ under β₁ ≠ β₂. Therefore, (α₁ − α₂)(β₁ − β₂) < 0 is the condition. We found that this condition applies when the relationship between different uncertainties s_p1 and s_p2 and constant external noise s_l is as follows:

$\begin{array}{l}{s_p}_1{s_p}_2>{s_l}^2\end{array}$

The authors apologize for these errors, which does not change the scientific conclusions of the article in any way. The original article has been updated.