Closed hihunjin closed 3 years ago
This image is from your paper. You said "Via simple integration, ...". I did it by myself but the constant term $\frac{1}{2-t}$ is ambiguous. How did you get this term? What constraint did you apply?
We set the constant such that the limit t -> 2 exists and gives:
F_2(y) = sum_i (y_i - log y_i - 1)
This convex function induces the Itakura-Saito (aka Burg) divergence.
This image is from your paper. You said "Via simple integration, ...". I did it by myself but the constant term $\frac{1}{2-t}$ is ambiguous. How did you get this term? What constraint did you apply?