def shuffle_weight(path):
# path = "/Users/clemens/Documents/Code/ModelClemens/output/11_29-01_52_white_circle_l_vs_r_ALS_smallSTDP/initial"
network_architecture = dict(num_inh_neurons_per_layer=32 * 32,
num_exc_neurons_per_layer=64 * 64,
num_layers=4)
synapses = data.load_network(path, True, True)
new_weights = synapse_analysis.shuffle_weights_within_layer(
synapses, network_architecture, 128**2)
raw_new_weights = new_weights.weights.values
raw_new_weights.tofile("{}/shuffled_weights.bin".format(path))
def test_paths_to_neuron():
path = "/Users/clemens/Documents/Code/ModelClemens/output/11_29-01_52_white_circle_l_vs_r_ALS_smallSTDP/initial"
network_architecture = dict(num_inh_neurons_per_layer=32 * 32,
num_exc_neurons_per_layer=64 * 64,
num_layers=4)
synapses = data.load_network(path, True, True)
mask = synapse_analysis.Synapse_Mask(network_architecture, synapses)
filter_path = "/Users/clemens/Documents/Code/ModelClemens/Data/MatlabGaborFilter/centered_inputs/Filtered"
all_filter = data.load_filter_all_obj(filter_path)
bla = synapse_analysis.paths_to_neurons([-3485],
synapses,
0.9,
max_path_length=3)
return bla
def test_masks(self):
path = "/Users/clemens/Documents/Code/ModelClemens/output/10_26-18_14_short_test_loc1/initial"
synapses = data.load_network(path, True, True)
mask = synapse_analyis.Synapse_Mask(
dict(num_exc_neurons_per_layer=64 * 64,
num_inh_neurons_per_layer=32 * 32,
num_layers=4), synapses)
exc_FF = mask.exc_feed_forward()
exc_FB = mask.exc_feed_back()
exc_L = mask.exc_lateral()
inh_L = mask.inh_lateral()
exc_to_inh_L = mask.exc_to_inhibitory()
unique = exc_FF ^ exc_FB ^ exc_L ^ inh_L ^ exc_to_inh_L
assert (np.all(unique))
def test_trace_to_gabor():
path = "/Users/clemens/Documents/Code/ModelClemens/output/01_19-18_02_rounded_trained_after_random_pretraining/multi_e275"
network_architecture = dict(num_inh_neurons_per_layer=32 * 32,
num_exc_neurons_per_layer=64 * 64,
num_layers=4)
synapses = data.load_network(path, True, True)
mask = synapse_analysis.Synapse_Mask(network_architecture, synapses)
e2e_ff = mask.exc_feed_forward()
bt = synapse_analysis.BackTracer(
synapses,
network_architecture,
percentage_thresholds=[0.1, 0.1, 0.1, 0.1],
mask=e2e_ff)
result = bt.trace_back(3, 32, 32)
return result
def test_spike_correlations():
path = "/Users/clemens/Documents/Code/ModelClemens/output"
experiment = "01_11-15_00_long_test_with_trace"
extension = "trained_e285"
stimuli_ids = data.load_testing_stimuli_dict(path + "/" + experiment +
"/" + extension)
network_architecture = dict(num_exc_neurons_per_layer=64 * 64,
num_inh_neurons_per_layer=32 * 32,
num_layers=4)
neuron_mask = helper.NeuronMask(network_architecture)
synapses = data.load_network(path + "/" + experiment + "/" + extension)
spikes = data.load_spikes_from_subfolders(path, [experiment], [extension],
False)
seperated = helper.splitstimuli(spikes[0][0], 2)
selected = helper.take_multiple_elements_from_list(seperated,
stimuli_ids["1wcr"])
target_id = helper.position_to_id((1, 19, 10), True, network_architecture)
target_id = 9186
# pre_ids = neuron_mask.get_ids_of_random_neurons_of_type(200, neuron_mask.is_excitatory)
pre_ids = synapses.pre.values[synapses.post.values == target_id]
pre_ids = np.sort(np.unique(pre_ids))
normed_hist, times = poly.population_normalised_pre_spike_hist(
stimuli_spikes=selected,
target_neuron=target_id,
max_time_delay=0.2,
time_step=0.002,
network_architecture=network_architecture,
preneurons=pre_ids,
start_time=1.0,
multiple_spikes_per_bin_possible=True)
hist_new, times = poly.pre_spike_hist(
stimuli_spikes=selected,
target_neuron=target_id,
max_time_delay=0.2,
time_step=0.1,
preneurons=pre_ids,
start_time=1.0,
multiple_spikes_per_bin_possible=True)
hist_old, times = poly.pre_spike_hist(
stimuli_spikes=selected,
target_neuron=target_id,
max_time_delay=0.02,
time_step=0.1,
preneurons=pre_ids,
start_time=1.0,
multiple_spikes_per_bin_possible=False)
assert (np.all(hist_new >= hist_old))
print(hist_new.shape)