Dmitry Razdoburdin

> Hi, what is your decision about this PR?

Hi all, I have returned the check for column matrix initialization back. Unfortunately amount of reusable code becomes less in this case.

> Is the column matrix still initialized for approx tree method? Asking since as mentioned previously the construction of it is quite costly for approx since it happens every iteration....

Hello, everyone. I have started decomposition of this PR into the smaller parts. The first one is presented here: https://github.com/dmlc/xgboost/pull/7991

Dear @trivialfis , thanks for the review and comments. This block of changes is the part of the optimization proposed in https://github.com/dmlc/xgboost/pull/7208. @ShvetsKS is busy in other projects, so I...

Dear @trivialfis , it is a good idea. We use Telegram for conversations. I can add you to the chat if it is suitable for you, if not we can...

> @razdoburdin Thank you, please add @RAMitchell as well. I sent the invitation to the email from your github profiles

Hi @trivialfis , CI failure in the PR isn't related to PR contest: [Type and format checks for the Python package (ubuntu-latest)](https://github.com/dmlc/xgboost/actions/runs/9093726058/job/24993279923?pr=10269) 

> Hi, I updated the branch to fix other errors, please see the status of the latest run. I have fixed the build. All checks passed.

Hi @trivialfis , Checks failures are not related to the PR content: [Type and format checks for the Python package](https://github.com/dmlc/xgboost/actions/runs/8829052948/job/24239257499?pr=10222):  [Build wheel for macosx_x86_64](https://github.com/dmlc/xgboost/actions/runs/8829052947/job/24239255509?pr=10222):  [Build wheel for macosx_arm64](https://github.com/dmlc/xgboost/actions/runs/8829052947/job/24239256483?pr=10222):...