def test_multi_fid(t_array, params):
data = multi_fid(t_array=t_array, params=params)
fid_0 = fid(t_array, *params[0])
fid_1 = fid(t_array, *params[1])
assert fid_0 + fid_1 == pytest.approx(data)
def calculate_freq_domain_rmse(data, params_est, dwell):
N = len(data)
t_array, freq_array = gen_t_freq_arrays(N=N, dwell=dwell)
data_est = multi_fid(t_array=t_array, params=params_est)
data_est_fft = np.fft.fft(data_est) / np.sqrt(N)
data_fft = np.fft.fft(data) / np.sqrt(N)
return np.sqrt(mean_squared_error(data_fft.real, data_est_fft.real))
def test_llc_kbdm(t_array, data_brain_sim, dwell, params_brain_sim, N):
m_range = range(250, 260, 1)
results = llc_kbdm(
data=data_brain_sim,
dwell=dwell,
m_range=m_range,
p=1,
l=30,
)
line_list = results.line_list
line_list = filter_samples(line_list, amplitude_tol=1e-3)
assert len(line_list) == len(params_brain_sim)
est_data = multi_fid(t_array, line_list)
assert np.std(est_data.real - data_brain_sim.real) < 1e-3
assert np.std(est_data.imag - data_brain_sim.imag) < 1e-3
def data_brain_sim(df_params_brain_sim, t):
data = sig_gen.multi_fid(t, df_params_brain_sim.values)
return data
def iterative_llc_kbdm(data,
dwell,
m_range,
p=1,
l=None,
q=0.0,
max_iterations=5,
silhouette_threshold=0.6):
if max_iterations < 1:
raise ValueError("'max_iterations must be greater than zero")
curr_data_est = np.zeros_like(data)
line_lists = []
silhouettes = []
t_array, _ = gen_t_freq_arrays(N=len(data), dwell=dwell)
n_peaks = 0
silhouette_thresholds = np.linspace(silhouette_threshold, 0,
max_iterations)
for i in range(max_iterations):
print(f'Iteration #{i}')
curr_res = data - curr_data_est
results = llc_kbdm(data=curr_res,
dwell=dwell,
m_range=m_range,
p=p,
l=l,
q=q)
if len(results.line_list) == 0:
logging.info('No more peaks can be fitted. Stopping.')
break
filtered_index = np.nonzero((results.silhouette > np.percentile(
results.silhouette, silhouette_thresholds[i])))
line_list = results.line_list[filtered_index]
curr_data_est_i = multi_fid(t_array=t_array, params=line_list)
curr_data_est += curr_data_est_i
line_lists.append(line_list)
silhouettes.append(results.silhouette[filtered_index])
n_peaks += len(line_list)
print(f'Found {len(line_list)} peaks. Total: {n_peaks} peaks.')
line_lists = np.r_[line_lists]
line_list = np.concatenate(line_lists)
silhouettes = np.r_[silhouettes]
rmse = calculate_freq_domain_rmse(data=curr_data_est,
params_est=line_list,
dwell=dwell)
return IterativeLlcKbdmResult(line_list=line_list,
line_lists=line_lists,
silhouettes=silhouettes,
rmse=rmse)