# Settings
INCL_NEURONS = 'pass-QC'
INCL_SESSIONS = 'aligned-behavior'
ATLAS = 'beryl-atlas'
MIN_NEURONS = 1
PLOT = True
PRE_TIME = 0.6
POST_TIME = -0.1
TRIAL_CENTERS = np.arange(-10, 21, 3)
TRIAL_WIN = 5
BASELINE_TRIAL_WIN = 5
COLORS = (sns.color_palette('colorblind', as_cmap=True)[0],
sns.color_palette('colorblind', as_cmap=True)[3])
# Query session list
eids, probes = query_sessions(selection=INCL_SESSIONS)
results_df = pd.DataFrame()
for i in range(len(eids)):
print('\nProcessing session %d of %d' % (i+1, len(eids)))
# Load in data
eid = eids[i]
try:
spikes, clusters, channels = bbone.load_spike_sorting_with_channel(
eid, aligned=True, one=one)
ses_path = one.path_from_eid(eid)
trials = load_trials(eid)
except Exception as error_message:
print(error_message)
continue
VALIDATION = 'kfold-interleaved'
INCL_NEURONS = 'all' # all or no_drift
INCL_SESSIONS = 'aligned-behavior' # all, aligned, resolved, aligned-behavior or resolved-behavior
NUM_SPLITS = 5
CHANCE_LEVEL = 'pseudo-sessions' # pseudo-blocks, phase-rand, shuffle or none
ITERATIONS = 100 # for null distribution estimation
DATA_PATH, FIG_PATH, SAVE_PATH = paths()
FIG_PATH = join(FIG_PATH, 'WholeBrain')
DOWNLOAD_TRIALS = False
PRE_TIME = 0
POST_TIME = 0.3
# %% Initialize
# Query session list
sessions = query_sessions(selection=INCL_SESSIONS)
# Detect data format
if type(sessions) == pd.DataFrame:
ses_type = 'datajoint'
elif 'model' in sessions[0]:
ses_type = 'insertions'
else:
ses_type = 'sessions'
# Initialize decoder
clf = MultinomialNB()
def decode(pop_vector, trial_ids, num_splits, interleaved):
-1] = False # Exclude all rew. ommissions
if 'neg' in target:
incl_trials[trials['feedbackType'] == 1] = False # Exclude all rewards
if '0' in target:
incl_trials[
trials['signed_contrast'] != 0] = False # Only include 0% contrast
if ('prior' in target) and ('stim' in target):
incl_trials[
trials['signed_contrast'] != 0] = False # Only include 0% contrast
incl_trials[trials['reaction_times'] <
min_rt] = False # Exclude trials with fast rt
return incl_trials
# Query session list
eids, probes, subjects = query_sessions(selection=INCL_SESSIONS,
return_subjects=True)
# Load in all trials
if CHANCE_LEVEL == 'other-trials':
all_trials = pd.read_pickle(
join(SAVE_PATH, 'Ephys', 'Decoding', 'all_trials.p'))
all_trials = all_trials[get_incl_trials(all_trials, TARGET, EXCL_5050,
MIN_RT)] # trial selection
# %% MAIN
# Load in decoding done so far if required
if OVERWRITE:
decoding_result = pd.DataFrame(
columns=['subject', 'date', 'eid', 'probe', 'region'])
elif isfile(
"""
import numpy as np
import pandas as pd
from os.path import join
import matplotlib.pyplot as plt
from behavior_models import utils
from my_functions import query_sessions, paths, figure_style
from models.expSmoothing_stimside import expSmoothing_stimside as exp_stimside
from models.expSmoothing_prevAction import expSmoothing_prevAction as exp_prev_action
# Save path
_, fig_path, save_path = paths()
# Query all sessions that are aligned and meet behavioral criterion
eids, probes, subjects = query_sessions(selection='aligned-behavior',
return_subjects=True)
# Loop over subjects
results = pd.DataFrame()
for i, subject in enumerate(np.unique(subjects)):
print('\nStarting subject %s [%d of %d]\n' %
(subject, i + 1, len(np.unique(subjects))))
stimuli_arr, actions_arr, stim_sides_arr, pleft_arr, session_uuids = [], [], [], [], []
for j, eid in enumerate(eids[subjects == subject]):
data = utils.load_session(eid)
if data['choice'] is not None and data['probabilityLeft'][0] == 0.5:
stim_side, stimuli, actions, pLeft_oracle = utils.format_data(data)
stimuli_arr.append(stimuli)
actions_arr.append(actions)
stim_sides_arr.append(stim_side)
pleft_arr.append(pLeft_oracle)