def trial_split_half_RMs(trials, image_property, response_property, rng):
if rng is None:
rng = np.random.RandomState(0)
inds = rng.permutation(trials.shape[0])
half = inds.shape[0]/2
inds1, inds2 = inds[:half], inds[half:]
trials1 = trials[inds1]
trials2 = trials[inds2]
RM1 = CM.get_response_matrix(trials1, image_property, response_property, group_by_worker=True)
RM2 = CM.get_response_matrix(trials2, image_property, response_property, group_by_worker=True)
return RM1, RM2
def get_rms(data, split_field, image_property, response_property, split_field_vals=None):
RMs = []
if split_field is None:
RMs.append(cm.get_response_matrix(data, image_property, response_property, group_by_worker=True))
else:
if split_field_vals is None:
split_field_vals = np.unique(data[split_field])
for fval in split_field_vals:
rel_data = data[data[split_field] == fval]
RMs.append(cm.get_response_matrix(rel_data, image_property=image_property,
response_property=response_property, group_by_worker=True)[0])
return RMs
def composite_individual_self_consistency_all_metrics(trials, image_property, response_property):
# Get response matrices
RMs = []
trials_hash = hashlib.sha1(str(trials)).hexdigest()
for trial_array in trials:
response_matrix, _, _ = CM.get_response_matrix(trial_array, image_property,
response_property, condition=None,
group_by_worker=True)
RMs.append(response_matrix)
# What metrics can we apply to these response matrices?
if image_property == 'task_category' and response_property == 'Response':
metrics = small_cm_metrics+rm_metrics
if min([response_matrix.shape[0] for response_matrix in RMs]) > 2:
metrics += large_cm_metrics
else:
metrics = rm_metrics
metrics_results = {}
for metric in metrics:
#filename = image_property+response_property+metric
#if os.path.exists(filename):
# return cPickle.load(open(filename, 'rb'))
#else:
m, sc, me, sce = metrics_from_confusion_mat(RMs, metric)
metrics_results[metric] = {'metric_values': m, 'self_consistency': sc,
'metric_error': me, 'self_consistency_error': sce}
return {'trials_hash': trials_hash, 'response_property': response_property, 'image_property': image_property,
'metrics_results': metrics_results, 'consistency_type': 'composite_individual'}
def trial_split_half_RMs(trials, image_property, response_property, rng):
if rng is None:
rng = np.random.RandomState(0)
inds = rng.permutation(trials.shape[0])
half = inds.shape[0] / 2
inds1, inds2 = inds[:half], inds[half:]
trials1 = trials[inds1]
trials2 = trials[inds2]
RM1 = CM.get_response_matrix(trials1,
image_property,
response_property,
group_by_worker=True)
RM2 = CM.get_response_matrix(trials2,
image_property,
response_property,
group_by_worker=True)
return RM1, RM2
def get_rms(data,
split_field,
image_property,
response_property,
split_field_vals=None):
RMs = []
if split_field is None:
RMs.append(
cm.get_response_matrix(data,
image_property,
response_property,
group_by_worker=True))
else:
if split_field_vals is None:
split_field_vals = np.unique(data[split_field])
for fval in split_field_vals:
rel_data = data[data[split_field] == fval]
RMs.append(
cm.get_response_matrix(rel_data,
image_property=image_property,
response_property=response_property,
group_by_worker=True)[0])
return RMs
def composite_individual_self_consistency_all_metrics(trials, image_property,
response_property):
# Get response matrices
RMs = []
trials_hash = hashlib.sha1(str(trials)).hexdigest()
for trial_array in trials:
response_matrix, _, _ = CM.get_response_matrix(trial_array,
image_property,
response_property,
condition=None,
group_by_worker=True)
RMs.append(response_matrix)
# What metrics can we apply to these response matrices?
if image_property == 'task_category' and response_property == 'Response':
metrics = small_cm_metrics + rm_metrics
if min([response_matrix.shape[0] for response_matrix in RMs]) > 2:
metrics += large_cm_metrics
else:
metrics = rm_metrics
metrics_results = {}
for metric in metrics:
#filename = image_property+response_property+metric
#if os.path.exists(filename):
# return cPickle.load(open(filename, 'rb'))
#else:
m, sc, me, sce = metrics_from_confusion_mat(RMs, metric)
metrics_results[metric] = {
'metric_values': m,
'self_consistency': sc,
'metric_error': me,
'self_consistency_error': sce
}
return {
'trials_hash': trials_hash,
'response_property': response_property,
'image_property': image_property,
'metrics_results': metrics_results,
'consistency_type': 'composite_individual'
}
def test_off_diagonal():
RM, _, _ = CM.get_response_matrix(
CM.get_data('hvm_basic_categorization_new', 'category'),
'task_category', 'Response')
print u.symmetrize_confusion_matrix(RM, take='off_diagonal')
def test_off_diagonal():
RM, _, _ = CM.get_response_matrix(CM.get_data('hvm_basic_categorization_new', 'category'),
'task_category', 'Response')
print u.symmetrize_confusion_matrix(RM, take='off_diagonal')
