Exercise on aliceml

Optimization example on aliceml

Connect to aliceml

ssh -Y <yourusername>@aliceml

Create the virtual environment and download the software(to be done only once)

ml-create-virtualenv
ml-activate-virtualenv
ml-activate-alilocal
ml-activate-root

Now clone the repository in your home directory:

git clone https://github.com/ginnocen/MachineLearningHEP.git

Install the needed packages (do it from your home directory:

cd MachineLearningHEP
pip3 install -e .

Install additional packages not installed automatically:

pip3 install root-numpy

Load the virtual environment (to be done every thing you connect to the server)

  ml-activate-virtualenv

Adjust your particle database:

In data/database_ml_parameters_LcpK0spp.yml change the output folders name in line 287-288 to:

mlplot: mlplot # to be removed  
mlout: mlout # to be removed 

And reduce the number of background and signal candidates in line 273-274 to a smaller number for test:

nbkg: 5000
nsig: 5000

Set which step of the optimization you want to run:

In default_complete.yaml you can set them changing false to true in lines 28-42. A recommended set of steps are docorrelation, dotraining, docrossvalidation, dolearningcurve, doroc, doimportance