Maelstrom

class maelstrom.maelstrom.Maelstrom(time, flux, freq=None, name='', model=None, **kwargs)

Bases: maelstrom.maelstrom.BaseOrbitModel

The real deal. This class provides an orbit model for an arbitrarily sized binary system, where each frequency freq is assigned a separate lighttime (asini). The time delays are forward-modelled directly onto the light curve time and flux data.

Parameters

• time (array) – Time values of the light curve

• flux (array) – Flux values of the light curve

• freq (array, optional) – Frequencies on which to model the time

• If none are supplied, Maelstrom will attempt to find the (delay.) –

• optimal frequencies. Defaults to None. (most) –

• name (str, optional) – Model name. Defaults to ‘’.

optimize(vars=None, verbose=False, **kwargs)

pin_orbit_model(opt=None)

Pins the orbit model to attribute the frequencies to the correct stars. In doing so, the lighttimes are collapsed into a single value for each star.

Parameters

opt (dict) – Results of the optimize function for the model. If none is supplied, Maelstrom will optimize first.

Returns

Return type

PB2Model or PB1Model, both instances of a PyMC3.model.Model

setup_orbit_model(period=None, eccen=None)

Generates an unpinned orbit model for the system. Each frequency in the model will be assigned a lighttime (asini). Optimizing this model will reveal which frequencies, if any, are associated with which stars. In general, this model should not be sampled.

Parameters

• period (float) – Initial period of the orbital system. If none is supplied (default), Maelstrom will attempt to find a suitable period by directly examining the time delays.

• eccen (float) – Initial eccentricity of the system, defaults to 0.5 if None supplied.

class maelstrom.maelstrom.PB1Model(time, flux, freq=None, name='PB1', model=None)

Bases: maelstrom.maelstrom.BaseOrbitModel

init_orbit(period, asini, with_eccen=True, with_gp=False)

optimize(vars=None)

Optimises the model.

Parameters

vars (array of model parameters, optional) – parameters of the model to be optimized, by default None

Returns

optimisation results

Return type

dict

class maelstrom.maelstrom.BaseOrbitModel(time, flux, freq=None, name='', model=None, **kwargs)

Bases: pymc3.model.Model

evaluate(var, opt=None)

Convenience function which wraps exoplanet.utils.eval_in_model()

first_look(segment_size=None, save_path=None, **kwargs)

Shows the light curve, its amplitude spectrum, the time delay signal, and the periodogram of the time delay signal in one convenient function. This is useful if you want to check whether a star may be a PM binary. However, sometimes only the strongest peak in the star will show a TD signal.

Args: segment_size (float, optional): Segment size in which to

subdivide the light curve, in units of time. Defaults to None. save_path (str, optional): If you want to save the output of the axis, pass a save path. Defaults to None.

Returns: array: Array of matplotlib axes.

get_period_estimate(**kwargs)

Estimates the period from the segmented time delay

Returns

period – Estimate of the period from extracted time delays.

Return type

float

get_time_delay(segment_size=None)

Calculates the time delay signal, splitting the light curve into chunks of width segment_size. A smaller segment size will increase the scatter of the time delay signal, especially for low frequencies.

Parameters

segment_size (float) – Segment size in which to separate the light curve, in units of the light curve time. For example, the default segment size of 10 will separate a 1000 d long light curve in 100 segments of 10 d each.

Returns

• time_midpoint (numpy.ndarray) – Midpoints of time for each segment in the light curve

• time_delay (numpy.ndarray) – Values of the extracted time delay in each segment.

get_weights(norm=True)

Calculates the amplitudes of each frequency, returning an array of amplitudes. This is useful for calculating the weighted average time delay.

Returns

(potentially) normalised amplitudes of each frequency

Return type

weights

period_search(periods=None)

Optimizes a model over a grid of periods

Parameters

periods (Array-like, optional) – Grid of periods over which to optimize, by default None

plot_periodogram(ax=None)

Plots the periodogram of the light curve with the model frequencies overlaid.

Parameters

ax (matplotlib axis, optional) – do you want an axis or not, by default None

Returns

this is also an axis.

Return type

ax

plot_time_delay(ax=None, show_weighted=True, **kwargs)

Plots the time delay. **kwargs go into get_time_delay.

plot_time_delay_periodogram(ax=None, **kwargs)

Plots the time delay periodogram (i.e. the amplitude spectrum of the time delay signal)

Parameters

• ax (matplotlib axis, optional) – Axis on which to plot. If None is

• a new one will be generated. Defaults to None. (supplied,) –

Returns

Axis on which the periodogram is plotted.

Return type

matplotlib axis

plot_time_delay_periodogram_period(min_period=None, max_period=None, ax=None, annotate=True, return_res=False, **kwargs)

Plots the time delay periodogram

profile()

Profiles the current model, returning the runtime of each node in your Theano graph

Returns

[description]

Return type

[type]

sample(tune=3000, draws=3000, start=None, target_accept=0.9, **kwargs)

Samples the model using the exoplanet PyMC3 sampler. By default, this will sample from 2 chains over 2 cores simultaneously.

Parameters

• tune (float, optional) – Number of tuning steps for the sampler (default 3000)

• draws (float, optional) – Number of samples from which to populate the trace (default 3000)

• start (dict, optional) – Starting location of the sampler. If none is supplied, the sampler will first optimize the model.

• target_accept (float, optional) – The target acceptance ratio of the NUTS sampler (default 0.9).

• **kwargs – Keyword arguments to pass to sample.tune and sample.sample

Returns

trace

Return type

pm trace object?

to_eddy()