Module connectome.visualization.group_imp

functions to calculate the group feature importance

Functions

def group_only_permutation_FI(model, Xtest, ytest, groups_df, m=10)

group only permutation feature importance, see paper: Au, Quay, et al. "Grouped feature importance and combined features effect plot." arXiv preprint arXiv:2104.11688 (2021), section 2.2.2. Compare the metric after permuting all features jointly with he metric after permuting all features except the considered group. For more information on how to use this function within the framework, see the documentation of the viz_framework function.

Args

model

fitted model

Xtest

pd.Dataframe containing test data

ytest

pd.Series containing labels of test data

groups_df

pd.Dataframe with columns region (e.g. yeo7) and conn_name (names of connectivity matrix)

m

number of random shuffles per group, default 10

Returns

a pd.DataFrame containing the regions and the associated increase in MSE/decrease in accuracy

def grouped_permutation_FI(model, Xtest, ytest, groups_df, m=10)

grouped permutation feature importance, see paper: Au, Quay, et al. "Grouped feature importance and combined features effect plot." arXiv preprint arXiv:2104.11688 (2021), section 2.2.1. Shuffle observations of group g and calculate metric after shuffle. For more information on how to use this function within the framework, see the documentation of the viz_framework function.

Args

model

fitted model

Xtest

pd.Dataframe containing test data

ytest

pd.Series containing labels of test data

groups_df

pd.Dataframe with columns region (e.g. yeo7) and conn_name (names of connectivity matrix)

m

number of random shuffles per group, default 10

Returns

a pd.DataFrame containing the regions and the associated increase in MSE/decrease in accuracy if features from that group are shuffled