bydmitry / bydmitry.github.io Goto Github PK

Hello, and thanks for your great writeup of the Efron approximation at https://bydmitry.github.io/efron-tensorflow.html !

I think there may be a mistake in the arithmetic directly below the table:

i=5,m=1:0.2−log(1.22−0)=0.00114914

The 1.22 number should be exactly exp(0.2) (the last timestep has no ties). In fact the resulting number -9.8596 is different from what's computed below, -9.8594456. I think this results from truncation.