Stacking is a powerful machine learning technique, but handling an excessive amount of models can be painful. For example, after training the stacking models as below:
If we want to change only some parameters in lasso
, then only lasso
and svr
need to be retrained. However, to reuse the other models/transformations and their output, we need to manually save them.
What if a package can save and index every model that we have trained? What if a package can automatically search the old models we trained and fetch them whenever we need them back?
Check out the demo code of mlforest
right now!
A simple example of a usage of the package. First create a virtual environment with virtualenv
called demo_mlforest
. Then:
cd demo_mlforest/
source bin/activate
git clone https://github.com/LukeLinEx/mlforest.git
cd mlforest
pip install -e .
jupyter notebook