Closed sbuechler closed 3 weeks ago
In Section 10.1, log loss is defined as -log \hat{p_i}(y_i). This appears to simply take the log of the predicted probability of y_i. This doesn't include the truth in the calculation. Do you want to multiply that log with y_i ?
I now see this is correct. ~ sorry.
In Section 10.1, log loss is defined as -log \hat{p_i}(y_i). This appears to simply take the log of the predicted probability of y_i. This doesn't include the truth in the calculation. Do you want to multiply that log with y_i ?