Then we should pass custom_objective to XGBRegressor's objective parameter to train the model with our objective. Return gradient(predt, dtrain), hessian(predt, dtrain) Return hessian_clf(predt, dtrain) + hessian_reg(predt, dtrain) Return gradient_clf(predt, dtrain) + gradient_reg(predt, dtrain) XGBRegressor is a general purpose notebook for model training using XGBoost. Bulk of code from Complete Guide to Parameter Tuning in XGBoost.
If they are twice differentiable, we can implement their gradient and hessian functions and combine them into an overall objective: def gradient_clf(predt, dtrain): A simple implementation to regression problems using Python 2.7, scikit-learn, and XGBoost. Assume that we have classification and regression objectives.
#Python xgbregressor objective code how to
Here the documentation explains how to create custom objectives. I see one possible solution using xgboost (it supports scikit-learn API). Or do I need to have more than one model? How can I solve this kind of problem in machine learning with one model? I found a reasonable approach in sklearn, but I don't think it assumes the case where regression and classification are mixed. In the case of deep learning, we know that with Keras we can use approaches such as This problem is a mixture of regression and classification. It is to predict age and gender from mugshots using the UTKFace dataset. The closest example to the problem I am facing is the following. It was also pointed out to me that I need to be more specific about my problem, which I will explain below. I would like to do regression and classification prediction in one model in this way. In tensorflow, we can branch the architecture. We can use them to output only regression or only classification, but what if we want to output a combination of regression and classification? Do I need to create separate models? In sklearn, there are MultiOutputRegressor and MultiOutputClassifier.
0 kommentar(er)
