XGBoost error - Unknown objective function reg:squarederror

Question

I am training a xgboost model for regression task and I passed the following parameters -

params = {'eta':0.4, 'max_depth':5, 'colsample_bytree':0.6, 'objective':'reg:squarederror'}
num_round = 10
xgb_model = xgboost.train(params, dtrain_x, num_round)

In the training phase I get the following error-

XGBoostError: b'[18:03:23] C:\Users\xgboost\src\objective\objective.cc:23: Unknown objective function reg:squarederror'

While in the docs, it is clearly a valid objective function. Can anyone tell me why am I getting this error?

INFO- I am using python 3.7.3 on windows and xgboost version is 0.82

Did you try to upgrade xgboost (pip install --upgrade xgboost) and scipy (pip install --upgrade scipy)? It worked for me today. I had the same issue — Lumber Jack, May 21 '19 at 14:42
@LumberJack I just upgraded xgboost to version 0.90, it worked. Thanks. — Ankit Seth, May 22 '19 at 08:07

score 5 · Answer 1 · edited Oct 09 '20 at 08:40

5

xgb_model = xgboost.train(**params, dtrain_x, num_round)

This would work for all versions. It is the way to pass **kwargs as a dictionary.

What does ** (double star/asterisk) and * (star/asterisk) do for parameters?

edited Oct 09 '20 at 08:40

fcdt

2,371
5
14
26

answered Oct 09 '20 at 03:16

T.singh

71
2
6

2

This doesn't work in any version, since python doesn't allow keyword arguments before positional arguments. – Quaternion Dec 20 '20 at 13:44

XGBoost error - Unknown objective function reg:squarederror

1 Answers1