def quantile_description(quantiles, name, source):
"""
This should determine which cutpoint range each point
belongs to and designate that as its state.
Example:
If in our source, we have under error a value of 230 and
we determine that this is between 0.1 and 0.2 quantile, then
the state will be State(error=(0.1, 0.2)).
Args:
quantiles (List[float]): This is the list of desired quantiles or
cutpoints, e.g., [0.0, 0.1, 0.9, 1.0]
name (str): The name of the relevant value in the Source
source (Source): The Source of data, this is needed to compute the
quantiles
Returns:
StateDescription: The quantile description
"""
quants = source.get_quantiles(name, quantiles)
def get_quantile(value):
for i, (x1, x2) in enumerate(zip(quants, quants[1:])):
if x1 <= value <= x2:
return (quants[i], quants[i + 1])
return states.StateDescription(**{name: get_quantile})
def sample_bin_description(bin_width, name, distribution):
"""
This function will create a state description where the value
given by name is described by an interval of width bin_width.
The limits of the interval will always be multiples of the bin_width.
If bin_width is 10, and the value under name is 12,
then its state would be sampled from (10, 20).
Evaluating this state will involve sampling from a
distribution that is passed in.
Args:
bin_width (float): The desired width of the bin
name (str): The name of the relevant value
distribution (BaseDistribution): A distribution from which we sample
conditioned on being in the bin of the state.
Returns:
StateDescription: The bin description
"""
def get_bin_value(value):
floor_value = value - (value % bin_width)
return IntervalSampledValue(distribution, floor_value,
floor_value + bin_width)
return states.StateDescription(**{name: get_bin_value})
def sample_quantile_description(quantiles, name, source, distribution):
"""
This should determine which cutpoint range each point
belongs to and designate this as the state. This will have the
QuantileSampledValue associated with it from which evaluating it will
sample from the distribution passed in . The sample will have its
quantile in the state's interval.
Example:
If in our source, we have under error a value of 230 and
we determine that this is between 0.1 and 0.2 quantile, then
the state will be State(error=sample from (0.1, 0.2)).
Args:
quantiles (List[float]): This is the list of desired quantiles or
cutpoints, e.g., [0.0, 0.1, 0.9, 1.0]
name (str): The name of the relevant value in the Source
source (Source): The Source of data, this is needed to compute the
quantiles
distribution (BaseDistribution): A distribution fitted to the values
in the column of source
Returns:
StateDescription: The quantile description
"""
quants = source.get_quantiles(name, quantiles)
def get_quantile_value(value):
for i, (x1, x2) in enumerate(zip(quants, quants[1:])):
if x1 <= value <= x2:
lower, upper = (quantiles[i], quantiles[i + 1])
return QuantileSampledValue(distribution, lower, upper)
return states.StateDescription(**{name: get_quantile_value})
def identity_description(name):
"""
For a certain value, the state of that value will be exactly that
value.
Example:
If in our source, we have under derivative_pattern, we have (0,0,0),
then the state will be State(derivative_pattern=(0,0,0))
Args:
name (str): The relevant value
Returns:
StateDescription: The identity description
"""
def identity(x):
return x
return states.StateDescription(**{name: identity})
def bin_description(bin_width, name):
"""
This function will create a state description where the value
given by name is described by an interval of width bin_width.
The limits of the interval will always be multiples of the bin_width.
If bin_width is 10, and the value under name is 12,
then its state would be (10, 20).
Args:
bin_width (float): The desired width of the bin
name (str): The name of the relevant value
Returns:
StateDescription: The bin description
"""
def get_bin(value):
floor_value = value - (value % bin_width)
return floor_value, floor_value + bin_width
return states.StateDescription(**{name: get_bin})