| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101 |
- import numpy as np
- import scipy.stats as stats
- from scripts.interneuron_placement import get_excitatory_phases_in_inhibitory_axon, create_grid_of_excitatory_neurons, \
- Pickle
- def distance(pickle_1, pickle_2):
- return np.sqrt((pickle_1.x - pickle_2.x) ** 2 + (pickle_1.y - pickle_2.y) ** 2)
- def contains(pickle, XX, YY):
- d1, d2 = distance_to_points(pickle, XX, YY)
- return d1 + d2 < pickle.c
- def distance_to_points(pickle, XX, YY):
- x_p_1, y_p_1 = pickle.get_p1()
- x_p_2, y_p_2 = pickle.get_p2()
- d1 = np.sqrt((XX - x_p_1) ** 2 + (YY - y_p_1) ** 2)
- d2 = np.sqrt((XX - x_p_2) ** 2 + (YY - y_p_2) ** 2)
- return d1, d2
- def get_overlap(pickle_1, pickle_2, ds=0.1):
- if distance(pickle_1, pickle_2) > 2 * (max(pickle_1.a, pickle_1.b) + max(pickle_2.a, pickle_2.b)):
- overlap = 0
- else:
- bb_x_min = pickle_1.x - max(pickle_1.a, pickle_1.b)
- bb_x_max = pickle_1.x + max(pickle_1.a, pickle_1.b)
- bb_y_min = pickle_1.y - max(pickle_1.a, pickle_1.b)
- bb_y_max = pickle_1.y + max(pickle_1.a, pickle_1.b)
- bb_X, bb_Y = np.meshgrid(np.arange(bb_x_min, bb_x_max, ds), np.arange(bb_y_min, bb_y_max, ds))
- in_1 = contains(pickle_1, bb_X, bb_Y)
- in_2 = contains(pickle_2, bb_X, bb_Y)
- overlap = np.sum(np.logical_and(in_1, in_2)) * ds ** 2
- return overlap
- def get_uniform_grid(sheet_size, grid_points_per_row):
- XX, YY = np.meshgrid(np.linspace(0, sheet_size, grid_points_per_row),
- np.linspace(0, sheet_size, grid_points_per_row))
- return np.vstack((XX.reshape((np.prod(XX.shape),)), YY.reshape((np.prod(YY.shape),)))).T
- def get_overlap_matrix(axonal_clouds, ds):
- overlaps = np.zeros((len(axonal_clouds), len(axonal_clouds)))
- for row_idx, cloud1 in enumerate(axonal_clouds):
- for column_idx, cloud2 in enumerate(axonal_clouds[row_idx + 1:]):
- overlaps[row_idx, row_idx + 1 + column_idx] = get_overlap(cloud1, cloud2, ds)
- return overlaps
- def get_entropy(cloud, ex_positions, ex_tunings, entropy_bins):
- phase_vals = get_excitatory_phases_in_inhibitory_axon(ex_positions, ex_tunings, cloud)
- phase_distr, _ = np.histogram(phase_vals, entropy_bins, density=True)
- return stats.entropy(phase_distr)
- def get_entropies(axonal_clouds, ex_positions, ex_tunings, entropy_bins):
- return [get_entropy(cloud, ex_positions, ex_tunings, entropy_bins) for cloud in axonal_clouds]
- class SpatialLayout(object):
- def __init__(self, orientation_map, NE, NI, long_axis, short_axis, sheet_size):
- self.orientation_map = orientation_map
- self.NE = NE
- self.NI = NI
- self.long_axis = long_axis
- self.short_axis = short_axis
- self.sheet_size = sheet_size
- ex_positions, ex_tunings = create_grid_of_excitatory_neurons(sheet_size, sheet_size, int(np.sqrt(NE)),
- orientation_map)
- self.ex_positions = ex_positions
- self.ex_tunings = ex_tunings
- self.in_positions = get_uniform_grid(sheet_size, int(np.sqrt(NI)))
- def get_axonal_clouds(self, orientation_array):
- return [Pickle(in_x_y[0], in_x_y[1], self.long_axis, self.short_axis, orientation) for in_x_y, orientation in
- zip(
- self.in_positions, orientation_array)]
- def get_entropies(self, orientation_array, number_of_bins):
- entropy_bins = np.linspace(-np.pi, np.pi, number_of_bins + 1)
- axonal_clouds = self.get_axonal_clouds(orientation_array)
- return get_entropies(axonal_clouds, self.ex_positions, self.ex_tunings, entropy_bins)
- def get_mean_entropy(self, orientation_array, number_of_bins):
- return np.mean(self.get_entropies(orientation_array, number_of_bins))
- def get_negative_mean_entropy(self, orientation_array, number_of_bins):
- return -self.get_mean_entropy(orientation_array, number_of_bins)
- def get_overlap(self, orientation_array, ds):
- axonal_clouds = self.get_axonal_clouds(orientation_array)
- overlaps = get_overlap_matrix(axonal_clouds, ds)
- return np.sum(overlaps)
|
