columns='subject_nickname',
index=['signed_contrast'])
# COMPUTE THE ACROSS-MOUSE VARIANCE FOR EACH CONTRAST
# AVERAGE VARIANCE ACROSS CONTRASTS
choice_var = 1 / choicedat.var(axis=1).mean()
return choice_var
# ================================= #
# GET DATA FROM TRAINED ANIMALS
# ================================= #
# query sessions - BASIC TASK
use_sessions, _ = query_sessions_around_criterion(criterion='trained',
days_from_criterion=[0, 2],
as_dataframe=False,
force_cutoff=True)
trial_fields = ('trial_stim_contrast_left', 'trial_stim_contrast_right',
'trial_response_time', 'trial_stim_prob_left',
'trial_feedback_type', 'trial_stim_on_time', 'trial_response_choice')
# query trial data for sessions and subject name and lab info
trials = use_sessions.proj('task_protocol') * behavior.TrialSet.Trial.proj(*trial_fields)
subject_info = subject.Subject.proj('subject_nickname') * \
(subject.SubjectLab * reference.Lab).proj('institution_short')
# Fetch, join and sort data as a pandas DataFrame
behav = dj2pandas(trials.fetch(format='frame')
.join(subject_info.fetch(format='frame'))
.sort_values(by=['institution_short', 'subject_nickname',
# import wrappers etc
from ibl_pipeline import reference, subject, behavior
from dj_tools import plot_psychometric, dj2pandas
# INITIALIZE A FEW THINGS
seaborn_style()
figpath = figpath()
pal = group_colors()
institution_map, col_names = institution_map()
col_names = col_names[:-1]
# ================================= #
# GET DATA FROM TRAINED ANIMALS
# ================================= #
use_sessions, use_days = query_sessions_around_criterion(
criterion='trained', days_from_criterion=[2, 0], as_dataframe=False)
# restrict by list of dicts with uuids for these sessions
b = use_sessions * subject.Subject * subject.SubjectLab * reference.Lab * \
behavior.TrialSet.Trial
# reduce the size of the fetch
b2 = b.proj('institution_short', 'subject_nickname', 'task_protocol',
'trial_stim_contrast_left', 'trial_stim_contrast_right',
'trial_response_choice', 'task_protocol', 'trial_stim_prob_left',
'trial_feedback_type')
bdat = b2.fetch(
order_by=
'institution_short, subject_nickname, session_start_time, trial_id',
format='frame').reset_index()
behav = dj2pandas(bdat)
behav['institution_code'] = behav.institution_short.map(institution_map)
assert (~behav.empty)
import numpy as np
from scipy import stats
from os.path import join
import seaborn as sns
from paper_behavior_functions import (query_sessions_around_criterion,
seaborn_style, institution_map,
group_colors, figpath)
from dj_tools import dj2pandas, fit_psychfunc
from ibl_pipeline import behavior, subject, reference
import scikit_posthocs as sp
# Settings
fig_path = figpath()
# Query sessions
sessions = query_sessions_around_criterion(criterion='biased',
days_from_criterion=[0, 15])[0]
sessions = sessions * subject.Subject * subject.SubjectLab * reference.Lab
# Create dataframe with behavioral metrics of all mice
learned = pd.DataFrame(columns=[
'mouse', 'lab', 'perf_easy', 'threshold', 'bias', 'reaction_time',
'lapse_low', 'lapse_high', 'n_trials', 'n_sessions'
])
for i, nickname in enumerate(np.unique(sessions.fetch('subject_nickname'))):
if np.mod(i + 1, 10) == 0:
print('Loading data of subject %d of %d' %
(i + 1, len(np.unique(sessions.fetch('subject_nickname')))))
# Get only the trials of the 50/50 blocks
trials = (sessions * behavior.TrialSet.Trial
kf = LeaveOneOut()
y_pred = np.empty(len(labels), dtype='<U5')
for train_index, test_index in kf.split(data):
clf.fit(data[train_index], labels[train_index])
y_pred[test_index] = clf.predict(data[test_index])
f1 = f1_score(labels, y_pred, average='micro')
cm = confusion_matrix(labels, y_pred)
return f1, cm
# %% Query sessions
if QUERY is True:
from paper_behavior_functions import query_sessions_around_criterion
from ibl_pipeline import reference, subject, behavior
use_sessions, _ = query_sessions_around_criterion(criterion='biased',
days_from_criterion=[1, 3])
use_sessions = use_sessions & 'task_protocol LIKE "%biased%"' # only get biased sessions
b = (use_sessions * subject.Subject * subject.SubjectLab * reference.Lab
* behavior.TrialSet.Trial)
b2 = b.proj('institution_short', 'subject_nickname', 'task_protocol',
'trial_stim_contrast_left', 'trial_stim_contrast_right', 'trial_response_choice',
'task_protocol', 'trial_stim_prob_left', 'trial_feedback_type',
'trial_response_time', 'trial_stim_on_time', 'time_zone')
bdat = b2.fetch(order_by='institution_short, subject_nickname, session_start_time, trial_id',
format='frame').reset_index()
behav = dj2pandas(bdat)
behav['institution_code'] = behav.institution_short.map(institution_map()[0])
else:
behav = load_csv('Fig4.csv')
biased_fits = pd.DataFrame()
# progress bar
tqdm.pandas(desc="model fitting")
# whether to query data from DataJoint (True), or load from disk (False)
query = True
institution_map, col_names = institution_map()
# ========================================== #
#%% 1. LOAD DATA
# ========================================== #
# Query sessions: before and after full task was first introduced
if query is True:
use_sessions, _ = query_sessions_around_criterion(
criterion='biased',
days_from_criterion=[2, 3],
as_dataframe=False,
force_cutoff=True)
trial_fields = ('trial_stim_contrast_left', 'trial_stim_contrast_right',
'trial_response_time', 'trial_stim_prob_left',
'trial_feedback_type', 'trial_stim_on_time',
'trial_response_choice')
# query trial data for sessions and subject name and lab info
trials = use_sessions.proj('task_protocol') * behavior.TrialSet.Trial.proj(
*trial_fields)
subject_info = subject.Subject.proj('subject_nickname') * \
(subject.SubjectLab * reference.Lab).proj('institution_short')
# Fetch, join and sort data as a pandas DataFrame
y_pred = np.array([])
y_true = np.array([])
for train_index, test_index in kf.split(resp):
train_resp = resp[train_index]
test_resp = resp[test_index]
clf.fit(train_resp, [labels[j] for j in train_index])
y_pred = np.append(y_pred, clf.predict(test_resp))
y_true = np.append(y_true, [labels[j] for j in test_index])
f1 = f1_score(y_true, y_pred, labels=np.unique(labels), average='micro')
unique_labels, label_counts = np.unique(labels, return_counts=True)
cm = confusion_matrix(y_true, y_pred, labels=unique_labels)
return f1, cm
# %% query sessions
use_sessions = query_sessions_around_criterion(criterion='ephys',
days_from_criterion=[2, 0])[0]
b = (use_sessions * subject.Subject * subject.SubjectLab * reference.Lab *
behavior.TrialSet.Trial
& 'task_protocol LIKE "%biased%"')
# load data into pandas dataframe
b2 = b.proj('institution_short', 'subject_nickname', 'task_protocol',
'trial_stim_contrast_left', 'trial_stim_contrast_right',
'trial_response_choice', 'task_protocol', 'trial_stim_prob_left',
'trial_feedback_type', 'trial_response_time', 'trial_stim_on_time',
'time_zone')
bdat = b2.fetch(
order_by=
'institution_short, subject_nickname, session_start_time, trial_id',
format='frame').reset_index()
behav = dj2pandas(bdat)
import numpy as np
from scipy import stats
from os.path import join
import seaborn as sns
from paper_behavior_functions import (query_sessions_around_criterion,
seaborn_style, institution_map,
group_colors, figpath)
from dj_tools import dj2pandas, fit_psychfunc
from ibl_pipeline import behavior, subject, reference
import scikit_posthocs as sp
# Settings
fig_path = figpath()
# Query sessions
sessions = query_sessions_around_criterion(criterion='trained',
days_from_criterion=[2, 0])[0]
sessions = sessions * subject.Subject * subject.SubjectLab * reference.Lab
# Create dataframe with behavioral metrics of all mice
learned = pd.DataFrame(columns=[
'mouse', 'lab', 'perf_easy', 'n_trials', 'threshold', 'bias',
'reaction_time', 'lapse_low', 'lapse_high'
])
for i, nickname in enumerate(np.unique(sessions.fetch('subject_nickname'))):
if np.mod(i + 1, 10) == 0:
print('Loading data of subject %d of %d' %
(i + 1, len(np.unique(sessions.fetch('subject_nickname')))))
# Get the trials of the sessions around criterion
trials = (sessions * behavior.TrialSet.Trial
pal = group_colors()
institution_map, col_names = institution_map()
col_names = col_names[:-1]
# colors for biased blocks
cmap = sns.color_palette([[0.8984375, 0.37890625, 0.00390625], [0.3, 0.3, 0.3],
[0.3671875, 0.234375, 0.59765625]])
sns.set_palette(cmap)
# ================================= #
# GET DATA FROM TRAINED ANIMALS
# ================================= #
if QUERY is True:
# query sessions
use_sessions, _ = query_sessions_around_criterion(
criterion='ephys', days_from_criterion=[2, 0], force_cutoff=True)
use_sessions = use_sessions & 'task_protocol LIKE "%biased%"' # only get biased sessions
# restrict by list of dicts with uuids for these sessions
b = (use_sessions * subject.Subject * subject.SubjectLab * reference.Lab *
behavior.TrialSet.Trial)
# reduce the size of the fetch
b2 = b.proj('institution_short', 'subject_nickname', 'task_protocol',
'session_uuid', 'trial_stim_contrast_left',
'trial_stim_contrast_right', 'trial_response_choice',
'task_protocol', 'trial_stim_prob_left', 'trial_feedback_type',
'trial_response_time', 'trial_stim_on_time')
# construct pandas dataframe
bdat = b2.fetch(
'subject_nickname')['institution_short'].apply(list).str[0]
# Change lab name into lab number
training_time['lab_number'] = training_time.lab.map(institution_map)
training_time = training_time.sort_values('lab_number')
# Save to csv
training_time.to_csv(join('data', 'Fig2d.csv'))
# %%=============================== #
# FIGURE 3
# ================================= #
print('Starting figure 3..')
# query sessions
use_sessions, _ = query_sessions_around_criterion(criterion='trained',
days_from_criterion=[2, 0])
use_sessions = use_sessions & 'task_protocol LIKE "%training%"' # only get training sessions
# query all trials for these sessions, it's split in two because otherwise the query would become
# too big to handle in one go
b = (use_sessions * subject.Subject * subject.SubjectLab * reference.Lab *
behavior.TrialSet.Trial)
# reduce the size of the fetch
b2 = b.proj('institution_short', 'subject_nickname', 'task_protocol',
'session_uuid', 'trial_stim_contrast_left',
'trial_stim_contrast_right', 'trial_response_choice',
'task_protocol', 'trial_stim_prob_left', 'trial_feedback_type',
'trial_response_time', 'trial_stim_on_time')
# construct pandas dataframe
