def _generate_gaussian_prior(
grid: UE4Grid,
means: "list of 2 means",
covariance_matrix: "2x2 covariance matrix",
initial_belief_sum=0.5
) -> "Tuple(np.array normed prior probs, prior_dict":
'''A method that returns the normed z vector as well as the prior dict. Given a 2d vector of means, 2X2 covariance matrix, returns a 2D gaussian prior'''
#maybe should check all dimensions of data here
prior = {}
x, y = np.mgrid[0:grid.get_no_points_x() *
grid.get_lng_spacing():grid.get_lng_spacing(),
0:grid.get_no_points_y() *
grid.get_lat_spacing():grid.get_lat_spacing()]
pos = np.empty(x.shape + (2, ))
pos[:, :, 0] = x
pos[:, :, 1] = y
rv = multivariate_normal(means, covariance_matrix)
z = rv.pdf(pos)
normed_z = z * (initial_belief_sum / z.sum())
#normed_z = normed_z.astype(float)
#return likelihoods in dict with grid points
for x_value in [_[0] for _ in x]:
for y_value in y[0]:
prior[Vector3r(x_value,
y_value)] = float(normed_z[x_value][y_value])
#place prior into ordered numpy array
list_of_grid_points = []
for grid_loc in grid.get_grid_points():
list_of_grid_points.append(prior[grid_loc])
list_of_grid_points = np.array(list_of_grid_points)
numpy_prior = np.append(list_of_grid_points, 1 - list_of_grid_points.sum())
return numpy_prior
def plot_prior(grid: UE4Grid, prior):
fig = plt.figure()
x, y = np.mgrid[0:grid.get_no_points_x() *
grid.get_lng_spacing():grid.get_lng_spacing(),
0:grid.get_no_points_y() *
grid.get_lat_spacing():grid.get_lat_spacing()]
ax = fig.gca(projection='3d')
prior = prior[:-1]
z_lim = prior.max() * 1.02
ax.set_zlim3d(0, z_lim)
if prior.shape != (len(x), len(y)):
prior = prior.reshape(len(x), len(y))
ax.plot_wireframe(x, y, prior)
ax.set_xlabel("physical x-axis of grid")
ax.set_ylabel("physical y-axis of grid")
ax.set_zlabel("Initial prob. evidence present at grid location")
return fig