pipeline = 'dexterity+prediction+network'
path = "/media/robbis/DATA/meg/viviana-hcp/derivatives/pipeline-%s/" % (
pipeline)
dataframe = get_results_bids(path,
field_list=loaded_keywords,
pipeline=[pipeline],
scores=scores)
for score in scores:
score_key = 'score_' + score
for clf in np.unique(dataframe['estimator__clf']):
for t in np.unique(dataframe['task']):
df = filter_dataframe(dataframe,
estimator__clf=[clf],
fx=[score],
task=[t])
df = apply_function(
df,
keys=['nodes_1', 'nodes_2', 'band', 'estimator__clf', 'fx'],
attr=score_key,
fx=lambda x: np.vstack(x).mean())
grid = sns.FacetGrid(df, col="band")
grid.map(plot_heatmap, "nodes_1", "nodes_2", score_key)
fname = "/home/robbis/Dropbox/PhD/experiments/blp-hcp-viviana/figure-regression-%s.png" % (
t)
grid.savefig(fname, dpi=150)
from pyitab.results import get_results_bids, filter_dataframe, apply_function
import seaborn as sns
import h5py
path = "/media/robbis/DATA/meg/viviana-hcp/derivatives/pipeline-blp-dexterity-trial"
dataframe = get_results_bids(path,
field_list=['sample_slicer','estimator__clf', 'feature_slicer'],
pipeline=['raw'],
)
dataframe['networks'] = dataframe['nodes_1']+'+'+dataframe['nodes_2']
average_df = apply_function(dataframe,
keys=['nodes_1', 'nodes_2', 'band', 'task', 'estimator__clf'],
attr='score_accuracy',
fx=lambda x:np.vstack(x).mean())
grid = sns.FacetGrid(dataframe, col="band", row="task", palette="tab20c",
height=1.5)
grid.map(sns.barplot, "networks", 'score_accuracy')
def plot_heatmap(nodes1, nodes2, accuracy, **kwargs):
df = dict(n1=nodes1, n2=nodes2, a=accuracy)
df = pd.DataFrame(df)
pdf = df.pivot("n1", "n2", "a")
nz = np.nonzero(np.isnan(pdf.values))
pdf.values[nz] = pdf.values[nz[::-1]]
sns.heatmap(pdf, annot=True, cmap="RdBu", fmt=".2f", center=.5, vmax=.8)
markersize=10,
markeredgecolor=None,
palette='Dark2')
axes = grid.axes
score_key = 'score_' + score
for clf in ['Lasso(alpha=0.5)']:
score_df = filter_dataframe(dataframe,
fx=[score],
estimator__clf=[clf])
average_df_score = apply_function(
score_df,
keys=['band', 'task', 'target_transformer__attr'],
attr=score_key,
fx=lambda x: np.nanmean(np.vstack(x)))
perm_df = filter_dataframe(dataframe_perm,
fx=[score],
estimator__clf=[clf])
average_df = apply_function(perm_df,
keys=[
'band', 'task', 'perm',
'estimator__clf',
'target_transformer__attr'
],
attr=score_key,
fx=lambda x: np.nanmean(np.vstack(x)))