add a notebook using the bagging model
this demo does the same thing as the minimal_dashboard
, but with a different
model. this model bagging consists in training 30 neuroquery models with
randomly subsampled vocabularies and averaging their coefficients. it is a
standard trick to gain a bit of robustness. however with this model we cannot
decompose the prediction into smoothing + encoding anymore, and it is less
obvious how to get good estimates for the coefficients' variance so this model
is not meant to replace the one currently on neuroquery.org.
neurodash therefore seems the right place to create an online demo