def fit_gamma_samples(samples, num_points=500, bound=np.pi, fix_location=False, return_fitted_data=True, should_plot=True, normalise=True, debug=True):
'''
Fit a Gamma distribution on the samples, optionaly plotting the fit
'''
if fix_location:
fit_alpha, fit_loc, fit_beta = spst.gamma.fit(samples, floc=1e-18)
else:
fit_alpha, fit_loc, fit_beta = spst.gamma.fit(samples)
# x = np.linspace(samples.min()*1.5, samples.max()*1.5, 1000)
x = np.linspace(-bound, bound, num_points)
# dx = 2.*bound/(num_points-1.)
fitted_data = spst.gamma.pdf(x, fit_alpha, fit_loc, fit_beta)
if debug:
print "Alpha: %.3f, Location: %.3f, Beta: %.3f" % (fit_alpha, fit_loc, fit_beta)
if should_plot:
if normalise:
hist_angular_data(samples, norm='max', bins=num_points)
plt.plot(x, fitted_data/np.max(fitted_data), 'r')
else:
hist_angular_data(samples, bins=num_points)
plt.plot(x, fitted_data, 'r')
plt.show()
if return_fitted_data:
return dict(parameters=np.array([fit_alpha, fit_loc, fit_beta]), fitted_data=fitted_data)
else:
return np.array([fit_alpha, fit_loc, fit_beta])
def fit_vonmises_samples(samples, num_points=500, return_fitted_data=True, should_plot=True, normalise=True, debug=True):
'''
Fit a Von Mises distribution on samples, optionaly plotting the fit.
'''
fit_kappa, fit_mu, fit_scale = spst.vonmises.fit(samples, fscale=1.0)
# x = np.linspace(samples.min()*1.5, samples.max()*1.5, 1000)
x = np.linspace(-np.pi, np.pi, num_points)
# dx = 2.*np.pi/(num_points-1.)
fitted_data = spst.vonmises.pdf(x, fit_kappa, loc=fit_mu, scale=fit_scale)
if debug:
print "mu: %.3f, kappa: %.3f" % (fit_mu, fit_kappa)
if should_plot:
if normalise:
hist_angular_data(samples, norm='density', bins=num_points)
plt.plot(x, fitted_data, 'r')
else:
hist_angular_data(samples, bins=num_points)
plt.plot(x, fitted_data, 'r')
plt.show()
if return_fitted_data:
return dict(parameters=np.array([fit_mu, fit_kappa]), fitted_data=fitted_data, support=x)
else:
return np.array([fit_mu, fit_kappa])
def fit_gaussian_samples(samples, num_points=500, bound=np.pi, should_plot = True, return_fitted_data = True, normalise = True, debug = False):
"""
Fit a 1D Gaussian on the samples provided.
Plot the result if desired
"""
mean_fit = np.mean(samples)
std_fit = np.std(samples)
# x = np.linspace(samples.min()*1.5, samples.max()*1.5, 1000)
x = np.linspace(-bound, bound, num_points)
dx = np.diff(x)[0]
print mean_fit
print std_fit
fitted_data = spst.norm.pdf(x, mean_fit, std_fit)
if debug:
print "Mean: %.3f, Std: %.3f" % (mean_fit, std_fit)
if should_plot:
if normalise:
hist_angular_data(samples, norm='max', bins=num_points)
plt.plot(x, fitted_data/np.max(fitted_data), 'r')
else:
hist_angular_data(samples, bins=num_points)
plt.plot(x, fitted_data, 'r')
plt.show()
if return_fitted_data:
return dict(parameters=np.array([mean_fit, std_fit]), fitted_data=fitted_data)
else:
return np.array([mean_fit, std_fit])