def identify_neuron_types():
# get trials for a single dummy neuron
real_data = load_mouse_data()
dummy_session = real_data[3]
#dummy_session = get_data_from_a_dummy_sesion()
# this is the first neuron from the dummy session:
print('Analyze two example neurons first.')
neuron_index = 0 # we will try to decide whether this is a choice neuron
left_vs_right_accuracy, accuracy_left_vs_nogo, accuracy_right_vs_nogo = compare_prediction_accuracies_for_neuron(dummy_session, neuron_index)
print('accuracies (left vs right, left vs no go, right vs no go):' + str(left_vs_right_accuracy) + ' ' + str(accuracy_left_vs_nogo) + ' ' + str(accuracy_right_vs_nogo))
neuron_type = guess_type(left_vs_right_accuracy, accuracy_left_vs_nogo, accuracy_right_vs_nogo)
print ("neuron idx", neuron_index, neuron_type)
print('----------------------------------------------------------------------')
neuron_index = 55 # we will try to decide whether this is a choice neuron
left_vs_right_accuracy, accuracy_left_vs_nogo, accuracy_right_vs_nogo = compare_prediction_accuracies_for_neuron(dummy_session, neuron_index)
print('accuracies (left vs right, left vs no go, right vs no go):' + str(left_vs_right_accuracy) + ' ' + str(accuracy_left_vs_nogo) + ' ' + str(accuracy_right_vs_nogo))
neuron_type = guess_type(left_vs_right_accuracy, accuracy_left_vs_nogo, accuracy_right_vs_nogo)
print ("neuron idx", neuron_index, neuron_type)
print('Analyze all neurons from this session and check accuracy of predictions:')
number_of_neurons = dummy_session['spks'].shape[0]
neuron_type_guesses = []
for neuron_index in range(number_of_neurons):
left_vs_right_accuracy, accuracy_left_vs_nogo, accuracy_right_vs_nogo = compare_prediction_accuracies_for_neuron(dummy_session, neuron_index)
neuron_type = guess_type(left_vs_right_accuracy, accuracy_left_vs_nogo, accuracy_right_vs_nogo)
neuron_type_guesses.append(neuron_type)
compare_predicted_to_ground_truth_type(dummy_session['dummy_type'], neuron_type_guesses)
def train_test_split(train_size: float = 0.8):
""" split the dataset into training and test set """
# load dataset
data = load_mouse_data()
split_trial_indices = create_dataset_split_indices(data, train_size)
train_data = get_training_data(data, split_trial_indices)
test_data = get_test_data(data, split_trial_indices)
return train_data, test_data
def main():
all_data = load_data.load_mouse_data()
print("Data loaded")
# test_data = split_test_and_training_data.get_test_data(all_data)
# training_data = split_test_and_training_data.get_training_data(all_data)
train_data, test_data = split_test_and_training_data.train_test_split(
train_size=0.8)
train_data = dimensionality_reduction(train_data, 0.9)
def get_training_data_for_simulation():
"""
Load the training data. We need this to get the real behavioural data.
We will later replace the real firing data with simulated data.
"""
mouse_data = load_data.load_mouse_data()
training_data = split_test_and_training_data.get_training_data(mouse_data)
simulated_data = copy.deepcopy(training_data)
return simulated_data
def test_create_dataset_split_indices():
data = load_mouse_data()
train_size = 0.7
session_id = 3
trials_split_dict = create_dataset_split_indices(data, train_size)
total_trials_in_sessionid = data[session_id]['spks'].shape[1]
assert isinstance(trials_split_dict, dict)
split_found = trials_split_dict[session_id]['train'].shape[0]
split_percent = int(train_size * total_trials_in_sessionid)
assert (split_found == split_percent)
def process_on_real_data(session_id: int = 10, single_neuron_id_for_train=5, single_neuron_id_for_test=9):
# default train test split is set = 0.8
data = load_mouse_data()
train_data, test_data = train_test_split()
session_data = train_data[session_id]
single_neuron_session_train_data = session_data['spks'][single_neuron_id_for_train]
single_neuron_session_test_data = session_data['spks'][single_neuron_id_for_test]
left_vs_right = get_accuracy_predicting_left_vs_right(session_data, single_neuron_session_train_data)
print ("left_vs_right : {}".format(left_vs_right))
left_vs_nogo = get_accuracy_predicting_left_vs_no_go(session_data, single_neuron_session_train_data)
print ("left_vs_no_go : {}".format(left_vs_nogo))
right_vs_nogo = get_accuracy_predicting_right_vs_no_go(session_data, single_neuron_session_train_data)
print ("right_vs_no_go : {}".format(right_vs_nogo))
def test_get_test_data():
# check if it splits the same way every time
data = load_mouse_data()
session_id = 10
train_size = 0.8
trials_split_dict = create_dataset_split_indices(data, train_size)
test_data_1 = get_test_data(data, trials_split_dict)
test_data_2 = get_test_data(data, trials_split_dict)
assert np.allclose(test_data_1[session_id]['contrast_right'],
test_data_2[session_id]['contrast_right'],
rtol=1e-05,
atol=1e-08,
equal_nan=False)
def test_get_test_and_training_data():
data = load_mouse_data()
train_size = 0.7
session_id = 6
training_data, test_data = train_test_split(train_size)
num_of_trials_all_example = data[session_id]['spks'].shape[1]
num_of_trials_test_example = test_data[session_id]['spks'].shape[1]
num_of_trials_training_example = training_data[session_id]['spks'].shape[1]
assert num_of_trials_all_example == (num_of_trials_test_example +
num_of_trials_training_example)
num_of_trials_all_example = data[session_id]['contrast_right'].shape[0]
num_of_trials_test_example = test_data[session_id]['contrast_right'].shape[
0]
num_of_trials_training_example = training_data[session_id][
'contrast_right'].shape[0]
assert num_of_trials_all_example == (num_of_trials_test_example +
num_of_trials_training_example)
def get_real_data_from_session(session_id: int = 10):
data = load_mouse_data()
session_data = data[session_id]
return session_id