murefi.ode

class murefi.ode.BaseODEModel(parameter_names: tuple, independent_keys: tuple)

Bases: object

A dynamic model that uses ordinary differential equations.

Methods

dydt(y, t, ode_parameters)	First derivative of the transient variables.
predict_dataset(template, parameter_mapping, ...)	Simulates an experiment that is comparable to the Dataset template. Args: parameter_mapping (ParameterMapping): maps elements in [parameters] to replicates in the [template] template (Dataset): template that the prediction will be comparable with parameters (array-like): prediction parameters, can be symbolic.
predict_replicate(parameters, template)	Simulates an experiment that is comparable to the Replicate template.
solver(y0, t, ode_parameters)	Solves the dynamic system for all T timepoints in t.
solver_vectorized(y0, t, ode_parameters)	Solves the dynamic system for all T timepoints in t with many parameter sets.

abstract dydt(y, t, ode_parameters)

First derivative of the transient variables. Needs to be overridden by subclasses.

Args:

y (array): current state of the system (n_y0,) t (float): time since intial state ode_parameters (array): system parameters (n_ode_parameters,)

Returns:

array: change in y at time t

predict_dataset(template: Dataset, parameter_mapping: ParameterMapping, parameters: Sequence | dict)

Simulates an experiment that is comparable to the Dataset template. Args:

parameter_mapping (ParameterMapping):

maps elements in [parameters] to replicates in the [template]

template (Dataset):

template that the prediction will be comparable with

parameters (array-like):

prediction parameters, can be symbolic

Returns:

prediction (Dataset): prediction result

predict_replicate(parameters: Sequence, template: Replicate) → Replicate

Simulates an experiment that is comparable to the Replicate template. Supports symbolic prediction and vectorized prediction from a matrix of parameters.

Args:

parameters (array-like):

The [parameters] sequence must be a tuple, list or numpy.ndarray, with the elements being a concatenation of y0 and ode_parameters.

Symbolic prediction requires a (n_parameters,) parameter vector of type {tuple, list, numpy.ndarray}. Elements may be a mix of scalars and PyTensor tensor variables.

Prediction of distributions requires a (n_parameters,) parameter vector of type {tuple, list, numpy.ndarray}. Elements may be a mix of scalars and vectors, as long as all vectors have the same length.

template (Replicate):

template that the prediction will be comparable with

Returns:

pred (Replicate): prediction result (contains Timeseries with y being numpy.arrays or Tensors)

solver(y0, t, ode_parameters) → dict

Solves the dynamic system for all T timepoints in t. Uses scipy.integrate.odeint and self.dydt to solve the system.

Args:

y0 (array): initial states (n_y0,) t (array): timepoints of the solution ode_parameters (array): system parameters (n_ode_parameters,)

Returns:

y_hat (dict):

keys are the independent keys of the model values are model states of shape (T,)

solver_vectorized(y0, t, ode_parameters) → dict

Solves the dynamic system for all T timepoints in t with many parameter sets. Uses scipy.integrate.odeint and self.dydt to solve the system.

Args:

y0 (array): initial states (n_y0, n_sets) t (array): timepoints of the solution ode_parameters (array): system parameters (n_ode_parameters, n_sets)

Returns:

y_hat (dict):

keys are the independent keys of the model values are model states of shape (T, n_sets)