Skip to content

Commit

Permalink
Change of docstring style to GoogleDocs, Fix for Bug that appeared wh…
Browse files Browse the repository at this point in the history 
…en samling a value using bad minimum, maximum bounds
  • Loading branch information
@sibange
sibange committed Mar 14, 2024
1 parent 4dc1f2a commit 26db9a8
Showing 1 changed file with 61 additions and 35 deletions.
96 changes: 61 additions & 35 deletions mms_msg/sampling/utils/distribution_model.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -22,11 +22,13 @@ class DistributionModel:
def __init__(self, samples: Optional[List[Union[int, float]]] = None, bin_size: Union[int, float] = 100,
allow_negative_samples: bool = False):
"""
:param bin_size: size of the histogram bins
:param samples: (optional) list of samples that should be added
:param allow_negative_samples: (optional) Allowing negative values to be added to the model.
Disabled by default.
Args:
samples: (optional) list of samples that should be added
bin_size: (optional) size of the histogram bins
allow_negative_samples: (optional) Allowing negative values to be added to the model.
Disabled by default.
"""

self.n = 0
self._distribution_prob = None
self._bin_size = bin_size
Expand Down Expand Up @@ -69,9 +71,10 @@ def standard_deviation(self) -> float:
return self._standard_deviation

def clear(self) -> None:
""" Removes all samples from the model and resets the related statistical values
:return: None
"""
Removes all samples from the model and resets the related statistical values
"""

self._distribution_prob = None
self._min_value = None
self._max_value = None
Expand All @@ -80,9 +83,11 @@ def clear(self) -> None:
self._standard_deviation = None

def fit(self, samples: Union[List[Union[int, float]]]) -> None:
""" Fits the distribution model to a number of samples. Previously estimated values will be overwritten.
:param samples: Samples to which the model is fitted. The samples can be given as list or as set.
:return: None
"""
Fits the distribution model to a number of samples. Previously estimated values will be overwritten.
Args:
samples: Samples to which the model is fitted. The samples can be given as list or as set.
"""

if len(samples) == 0:
Expand Down Expand Up @@ -129,15 +134,19 @@ def sample_value(self, rng: Optional[np.random.random] = None, random_state: Opt
It is also possible to restrict the area to an interval from which a sample is drawn.
In this case, the distribution inside the interval is normalized to the probability 1 and then used for sampling.
:param rng: (optional) The numpy rng that should be used, the rng should generate a number in the interval [0,1)
If not set a new uniform rng is used.
:param random_state: (optional) Seed for the default random number generator.
If not set, no seed is used for the rng, so the samples are no reproducible.
:param sample_integer: (optional) When set to true, the sampled value is an integer, otherwise it is a float.
Default: True.
:param minimum_value: (optional) minimal value that should be sampled (including minimum_value)
:param maximum_value: (optional) maximum value that should be sampled (excluding maximum_value)
:return: sample according to the distribution Integer, when sample_integer is True.
Args:
rng: (optional) The numpy rng that should be used, the rng should generate a number in the interval [0,1)
If not set a new uniform rng is used.
random_state: (optional) Seed for the default random number generator.
If not set, no seed is used for the rng, so the samples are no reproducible.
sample_integer: (optional) When set to true, the sampled value is an integer, otherwise it is a float.
Default: True.
minimum_value: (optional) minimal value that should be sampled (including minimum_value)
maximum_value: (optional) maximum value that should be sampled (excluding maximum_value)
Returns: Sample according to the distribution Integer. Returns an integer when sample_integer is set to True,
otherwise returns a float.
"""

if rng is None:
Expand All @@ -146,6 +155,9 @@ def sample_value(self, rng: Optional[np.random.random] = None, random_state: Opt
if self.n == 0:
raise AssertionError("No samples has been added to the model. Sampling not possible.")

if minimum_value is not None and maximum_value is not None and minimum_value >= maximum_value:
raise ValueError('When given the maximum value must be greater than the minimum value.')

if minimum_value is None:
p_min = 0
else:
Expand All @@ -156,6 +168,10 @@ def sample_value(self, rng: Optional[np.random.random] = None, random_state: Opt
else:
p_s = self.get_cdf_value(maximum_value)-p_min

if p_s <= 0:
raise ValueError('The probability that an element is in the given boundaries is 0 according to the'
' underlying model.')

temp = p_min + rng.random()*p_s

for (val, prob) in self.distribution_prob:
Expand All @@ -175,8 +191,10 @@ def get_cdf_value(self, value: Union[int, float]) -> float:
Returns the value of the cumulative distribution function (cdf) for the given value.
In other words returns the probability that a random sample is smaller than value.
:param value: Value for which the cdf should be evaluated
:return: Output of the cdf function at the given value.
Args:
value: Value for which the cdf should be evaluated
Returns: Output of the cdf function at the given value.
"""

if value < self.min_value:
Expand All @@ -202,21 +220,25 @@ def __repr__(self):
ret += " Variance:" + str(self.variance)
return ret

def plot(self, show = False, fig = None, ax = None):
def plot(self, show: bool = False, fig=None, ax=None):
"""
Creates a plot of the distribution model using matplotlib and
returns a figure and axes with the corresponding plot.
@:param show: (optional) When set to True the figure is directly shown
@:param fig: (optional) Figure on which a new axes with the plot is created.
Will be overwritten when ax is given.
When not given and also ax is not provided the function creates a new figure
with one axes and uses this for the plot.
@:param ax: (optional) axes on which the plot is created, when not provided
the function creates a new axes on the figure, when also the figure is not provided
then the function creates a new figure with one axes and uses this for the plot.
:return: Figure and axes with the plot of the distribution.
When an axis but no figure is given as input then the tuple (None,ax) is returned.
returns a figure and axes with the corresponding plot.
Args:
show: (optional) When set to True the figure is directly shown
fig: (optional) Figure on which a new axes with the plot is created.
Will be overwritten when ax is given.
When not given and also ax is not provided the function creates a new figure
with one axes and uses this for the plot.
ax: (optional) axes on which the plot is created, when not provided
the function creates a new axes on the figure, when also the figure is not provided
then the function creates a new figure with one axes and uses this for the plot.
Returns: Figure and axes with the plot of the distribution.
When an axis but no figure is given as input then the tuple (None,ax) is returned.
"""

import matplotlib.pyplot as plt

if self.n == 0:
Expand Down Expand Up @@ -245,10 +267,14 @@ def statistical_distance(d1: DistributionModel, d2: DistributionModel) -> float:
Calculates the statistical distance (total variation distance,
https://en.wikipedia.org/wiki/Total_variation_distance_of_probability_measures)
of two distribution models (d1 and d2).
:param d1: DistributionModel for comparison
:param d2: DistributionModel for comparison
:return: statistical distance
Args:
d1: DistributionModel for comparison
d2: DistributionModel for comparison
Returns: statistical distance
"""

if d1.n == 0:
raise AssertionError("No samples has been added to the first model. No comparison possible.")
elif d2.n == 0:
Expand Down

0 comments on commit 26db9a8

Please sign in to comment.