import warnings
import numpy as np
[docs]class Predictor():
""" Object representing a function to fit a learning curve (See :class:`learning_curves.LearningCurve`). """
def __init__(self, name, func, guess, inv=None, diverging=False, bounds=None):
""" Create a Predictor.
Args:
name (str): name of the function
func (Callable): lambda expression, function to fit
guess (Tuple): Initial parameters
inv (Callable): lambda expression corresponding to the inverse function.
diverging (bool): False if the function converge. In this case the first parameter must be the convergence parameter (enforced to be in [-inf,1]).
bounds (array of tuples): Bounds of the parameters. Default is [-inf, inf] for all parameters, except for the convergence parameter whose bounds are [-inf,1]
if diverging is True.
"""
self.name = name
self.func = func
self.guess = guess
self.params = self.guess
self.score = None
self.cov = {}
self.diverging = diverging
self.params_up = None
self.params_low = None
if callable(inv):
self.inv = lambda x, *args: inv(x, *args) if len(args) > 0 else inv(x, *self.params)
else:
self.inv = None
if bounds:
self.bounds = bounds
else:
self.bounds = (-np.inf, np.inf) if self.diverging else ([-np.inf] * (len(self.params)), [1]+[np.inf] * (len(self.params) - 1))
def __call__(self, x, *args):
# with warnings.catch_warnings():
#warnings.simplefilter("ignore", RuntimeWarning)
x = np.array(x) # Enforce x to be a np array because a list of floats would throw a TypeError
return self.func(x, *args) if len(args) > 1 else self.func(x, *self.params)
def __repr__(self):
return f"Predictor {self.name} with params {self.params} and score {self.score}"
[docs] def get_saturation(self):
""" Compute the saturation accuracy of the Predictor.
The saturation accuracy is the best accuracy you will get from the model without changing any other parameter than the training set size.
If the Predictor is diverging, this value should be disregarded, being meaningless.
Returns:
float: saturation accuracy of the Predictor.
This value is 1 if the Predictor is diverging without inverse function.
This valus is the first parameter of the Predictor if it is converging.
This value is calculated if the Predictor is diverging with inverse function.
"""
if not self.diverging: sat_acc = self.params[0]
elif callable(self.inv):
sat_acc = 1 # if predictor is diverging, set saturation accuracy to 1
sat_val = self.inv(sat_acc)
while not np.isfinite(sat_val): # Decrease the saturation accuracy until finding a value that is not inf
sat_acc -= 0.01
sat_val = self.inv(sat_acc)
else: sat_acc = 1 # Default value if diverging Perdictor
return sat_acc
def __eq__(self, other):
if not isinstance(other, Predictor): return RuntimeError("Trying to compare Predictor with not Predictor object.")
return self.name == other.name