Improving Concordance Index in Regression-based Survival Analysis: Evolutionary Discovery of Loss Function for Neural Networks
Will be available soon
For each dataset (Nwtco, Flchain, Support), there are 6 notebooks. Run each of them to do the following:
-
Experiment to optimize the full function f(x)+g(x), repeated 10 times.
-
Experiment to optimize the censored part g(x) and fixing the events part to f(x)=x^2, repeated 10 times..
-
Experiment to optimize the events part f(x) and fixing the censored part to g(x)=max(0,x)^2, repeated 10 times..
-
Comparison between the results of the optimization of the Full function f(x)+g(x), fixing g(x), and fixing f(x). You need to copy the results from the first three notebooks.
-
Comparison between MSCEsp and the truncated MSCEsp.
-
Comparison between optimization and MSCEsp fucntion. . You need to copy the results from the previous (1, 2, 3, and 5) notebooks.