# -*- coding: utf-8 -*-
# Simple Gaussian ModelLink
# Compatible with Python 3.5+
"""
GaussianLink
============
Provides the definition of the :class:`~GaussianLink` class.
"""
# %% IMPORTS
# Package imports
import numpy as np
# PRISM imports
from prism._internal import check_vals
from prism.modellink._modellink import ModelLink
# All declaration
__all__ = ['GaussianLink']
# %% CLASS DEFINITION
[docs]class GaussianLink(ModelLink):
"""
:class:`~prism.modellink.ModelLink` class wrapper for a simple Gaussian
model, used for testing the functionality of the *PRISM* pipeline.
Formatting data_idx
-------------------
x : float
The value that needs to be used for :math:`x` in the function
:math:`\\sum_i A_i\\exp\\left(-\\frac{(x-B_i)^2}{2C_i^2}\\right)` to
obtain the data value.
"""
[docs] def __init__(self, n_gaussians=1, *args, **kwargs):
"""
Initialize an instance of the :class:`~GaussianLink` class.
Optional
--------
n_gaussians : int. Default: 1
The number of Gaussians to use for the Gaussian model in this
instance. The resulting number of model parameters
:attr:`~prism.modellink.ModelLink.n_par` will be
:math:`3*n_{gaussians}`.
"""
# Set the number of Gaussians
self._n_gaussians = check_vals(n_gaussians, 'n_gaussians', 'pos',
'int')
# Set the name of this GaussianLink instance
self.name = 'GaussianLink_n%i' % (self._n_gaussians)
# Request single or multi model calls
self.call_type = 'hybrid'
# Request only controller calls
self.MPI_call = False
# Inheriting ModelLink __init__()
super().__init__(*args, **kwargs)
# %% GAUSSIANLINK CLASS PROPERTIES
@property
def n_gaussians(self):
"""
int: Number of Gaussians used in this :obj:`~GaussianLink` instance.
"""
return(self._n_gaussians)
# %% GAUSSIANLINK CLASS METHODS
def get_str_repr(self):
return(['n_gaussians=%r' % (self._n_gaussians)] if(
self._n_gaussians != 1) else [])
def get_default_model_parameters(self):
# Set default parameters for every Gaussian
A = [1, 10, 5]
B = [0, 10, 5]
C = [0, 5, 2]
# Create default parameters dict and return it
par_dict = {}
for i in range(1, self._n_gaussians+1):
par_dict['A%i' % (i)] = list(A)
par_dict['B%i' % (i)] = list(B)
par_dict['C%i' % (i)] = list(C)
return(par_dict)
def call_model(self, emul_i, par_set, data_idx):
par = par_set
mod_set = np.zeros([len(data_idx), *np.shape(par['A1'])])
for i, idx in enumerate(data_idx):
for j in range(1, self._n_gaussians+1):
mod_set[i] +=\
par['A%i' % (j)]*np.exp(-1*((idx-par['B%i' % (j)])**2 /
(2*par['C%i' % (j)]**2)))
return(mod_set.T)
def get_md_var(self, emul_i, par_set, data_idx):
return(0.01*np.ones_like(data_idx))