In [1]:

    

%matplotlib inline
import numpy as np
import matplotlib.pyplot as plt
import seaborn; seaborn.set()

from multiband_LS import LombScargle, SuperSmoother


    

In [2]:

    

def plot_width(period_true=0.5, T=1000, N=60, dy=0.01):
    rng = np.random.RandomState(0)
    omega_true = 2 * np.pi / period_true
    width_guess = 2 * np.pi / T
    
    t = T * rng.rand(N)
    y = np.sin(omega_true * t) + dy * rng.randn(N)
    
    offset = np.linspace(-4, 4, 100)
    sample_offset = 0.25 * (np.random.rand() + np.arange(-10, 10))
    samples = omega_true + sample_offset * width_guess
    
    omega_test = omega_true + width_guess * offset
    score = LombScargle().fit(t, y, dy).score(2 * np.pi / omega_test)
    sample_score = LombScargle().fit(t, y, dy).score(2 * np.pi / samples)
    
    plt.plot(offset, score)
    plt.plot(sample_offset, sample_score, 'ok')


    

In [3]:

    

from IPython.html.widgets import interact


    

In [4]:

    

interact(plot_width, period_true=(0.01, 1.0), T=(1, 2000), N=(10, 100), dy=(0.01, 1.0));


    

In [5]:

    

from multiband_LS import LombScargleAstroML


    

In [13]:

    

P = 0.6
N = 30
rng = np.random.RandomState(42)
t = rng.randint(0, 100, N) + 0.1 * rng.randn(N)
dy = 0.5
y = dy * np.sin(2 * np.pi * t / P)

model = LombScargle().fit(t, y, dy)

periods = np.linspace(0.2, 1.2, 5000)
scores = model.score(periods)

best_periods, best_scores = model.optimizer.find_best_periods(model, 5, return_scores=True)

plt.plot(periods, scores);
plt.plot(best_periods, best_scores, 'ok')
plt.ylim(0, 1.1);


    

    




Finding optimal frequency:
 - Using omega_step = 0.01281
 - Computing periods at 2045 steps from 0.20 to 1.20
Zooming-in on 10 candidate peaks:
 - Computing periods at 2010 steps






    

In [6]: