Creating a better validation set when test examples differ from training examples
Python
Latest commit a2e4ad1 May 17, 2016 @zygmuntz initial commit ⚡️
Permalink
Failed to load latest commit information.
numerai
santander
LICENSE initial commit ⚡️ May 17, 2016
README.md

README.md

Adversarial validation

The santander dir holds the scripts for the Santander competition:

distinguish_train_test.py - try to distinguish train/test set examples
validate.py - get validation AUC scores for logistic regression and random forest
predict.py - output test predictions from logistic regression and random forest

Similarly, the 'numerai' dir contains the Numerai scripts:

distinguish_train_test.py - try to distinguish train/test set examples
sort_train.py - sort training examples by their similarity to test examples
validate_sorted.py - get validation scores using for most test-like examples
predict.py - output test predictions