Closed kgourgou closed 9 years ago
The metric used, logarithmic-loss, is very common in those kind of competitions. It's root is in information theory, where it can be derived by the cross (or joint) entropy between two distributions.
http://en.wikipedia.org/wiki/Cross_entropy#Cross-entropy_error_function_and_logistic_regression
The metric used, logarithmic-loss, is very common in those kind of competitions. It's root is in information theory, where it can be derived by the cross (or joint) entropy between two distributions.
http://en.wikipedia.org/wiki/Cross_entropy#Cross-entropy_error_function_and_logistic_regression