Rerun the machine learning pipeline for channel power at alpha with model selection

[yacineMahdid]

In a previous task we used only the linear svm with a hard set parameter C. To have a more robust machine learning pipeline we need to not put a bias on which model we land on. The only criteria for the model is that they should be interpretable, but the parameter that the final model has or the type of model should be driven by the data.

At this point it would be a good idea to write down a report on what I've learned and what the current method looks like

[yacineMahdid]

This is solved by GridSearch