I see that with the xgboost library, we can tell the training process that some features are necessarily monotonic with the model's output - https://xgboost.readthedocs.io/en/latest/tutorials/monotonic.html. This forces the training procedure to avoid exploring models where there isn't a monotonic relationship between the given features and the output, which saves resources and improves accuracy. I wonder if such an option is available with neural nets? Plus, are there any papers testing improved accuracy of using this (with neural nets or other models)?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.