jpt.variables

© Copyright 2021, Mareike Picklum, Daniel Nyga.

Attributes

INVERT_IMPURITY

Classes

Variable	Abstract class for a variable name along with its
NumericVariable	Represents a continuous variable.
IntegerVariable	Represents an integer-valued variable.
SymbolicVariable	Represents a symbolic variable.
VariableMap	Convenience class for mapping a Variable object
VariableAssignment	Specialization of a VariableMap that maps a set
LabelAssignment	Maps a set of variables to values represented by
ValueAssignment	Maps a set of variables to values represented by

Functions

infer_from_dataframe(df[, scale_numeric_types, ...])

Creates the Variable instances from column types

Module Contents

class jpt.variables.Variable(name: str, domain: type[jpt.distributions.Distribution] | None = None, **settings)

Abstract class for a variable name along with its distribution class type.

Parameters:

• name – name of the variable

• domain – the class type (not an instance!) of the represented Distribution

MIN_IMPURITY_IMPROVEMENT = 'min_impurity_improvement'

SETTINGS

_name

_domain = None

settings

__getattr__(name)

property name: str

property domain

distribution() → jpt.distributions.Distribution

Create and return a new instance of the distribution type attached to this variable.

__str__()

__repr__()

__eq__(other)

__hash__()

property symbolic: bool

property numeric: bool

property integer: bool

abstract str(assignment, **kwargs) → str

to_json() → dict[str, Any]

static from_json(data: dict[str, Any]) → NumericVariable | SymbolicVariable | IntegerVariable

__getstate__()

__setstate__(state)

copy()

abstract assignment2set(assignment: Any)

Return a canonical representation of the variable assignment as a set in the corresponding type of set.

For a NumericVariable, a scalar assignment will be converted to a ContinuousSet instance, for a SymbolicVariable, a single value will be converted to a set collection.

If assignment is already in its canonical set representation, it will not be modified and returned as passed.

class jpt.variables.NumericVariable(name: str, domain: type[jpt.distributions.Numeric] | None = Numeric, min_impurity_improvement: float | None = None, blur: float | None = None, max_std: float | None = None, precision: float | None = None)

Bases: Variable

Represents a continuous variable.

Parameters:

• name – name of the variable

• domain – the class type (not an instance!) of the represented Distribution

BLUR = 'blur'

MAX_STDEV = 'max_std_lbl'

PRECISION = 'precision'

SETTINGS

to_json() → dict[str, Any]

static from_json(data: dict[str, Any]) → NumericVariable

property _max_std

property max_std

str(assignment: list | set | numbers.Number | jpt.base.intervals.NumberSet, **kwargs) → str

Construct a pretty-formatted string representation of the respective variable assignment.

Parameters:

• assignment – the value(s) assigned to this variable.

• fmt – [“set” | “logic”] use either set or logical notation.

• precision – (int) the number of decimals to use for rounding.

assignment2set(assignment: float | jpt.base.intervals.NumberSet) → jpt.base.intervals.NumberSet

Return a canonical representation of the variable assignment as a set in the corresponding type of set.

For a NumericVariable, a scalar assignment will be converted to a ContinuousSet instance, for a SymbolicVariable, a single value will be converted to a set collection.

If assignment is already in its canonical set representation, it will not be modified and returned as passed.

class jpt.variables.IntegerVariable(name: str, domain: type[jpt.distributions.Integer] | None, min_impurity_improvement: float | None = None)

Bases: Variable

Represents an integer-valued variable.

Parameters:

• name – name of the variable

• domain – the class type (not an instance!) of the represented Distribution

str(assignment, **kwargs) → str

_intset_logic_str(i: jpt.base.intervals.IntSet) → str

assignment2set(assignment: int | jpt.base.intervals.IntSet) → jpt.base.intervals.IntSet

Return a canonical representation of the variable assignment as a set in the corresponding type of set.

For a NumericVariable, a scalar assignment will be converted to a ContinuousSet instance, for a SymbolicVariable, a single value will be converted to a set collection.

If assignment is already in its canonical set representation, it will not be modified and returned as passed.

static from_json(data: dict[str, Any]) → IntegerVariable

to_json() → dict[str, Any]

jpt.variables.INVERT_IMPURITY = 'invert_impurity'

class jpt.variables.SymbolicVariable(name: str, domain: type[jpt.distributions.Multinomial] | None, min_impurity_improvement: float | None = None, invert_impurity: bool | None = None)

Bases: Variable

Represents a symbolic variable.

Parameters:

• name – name of the variable

• domain – the class type (not an instance!) of the represented Distribution

SETTINGS

static from_json(data: dict[str, Any]) → SymbolicVariable

to_json() → dict[str, Any]

str(assignment: set | numbers.Number, **kwargs) → str

assignment2set(assignment: Any)

Return a canonical representation of the variable assignment as a set in the corresponding type of set.

For a NumericVariable, a scalar assignment will be converted to a ContinuousSet instance, for a SymbolicVariable, a single value will be converted to a set collection.

If assignment is already in its canonical set representation, it will not be modified and returned as passed.

jpt.variables.infer_from_dataframe(df, scale_numeric_types: bool = True, min_impurity_improvement: float | None = None, blur: float | None = None, max_std: float | None = None, precision: float | None = None, unique_domain_names: bool = False, excluded_columns: dict[str, type] | None = None, remove_nan: bool = False)

Creates the Variable instances from column types in a Pandas or Spark data frame.

Parameters:

• df – the data frame object to generate the variables from.

• scale_numeric_types – whether or not to use scaled types for the numeric variables.

• min_impurity_improvement – the minimum improvement that a split must induce to be acceptable.

• blur – blur parameter for numeric variables.

• max_std – maximum standard deviation.

• precision – precision in [0, 1].

• unique_domain_names – for multiple calls of infer_from_dataframe containing duplicate column names the generated domain names will be unique.

• excluded_columns – user-provided domains for specific columns.

• remove_nan – skip all None or NaN or Inf values in the data to construct the numeric variable domains.

class jpt.variables.VariableMap(data: list[tuple] | dict = None, variables: collections.abc.Iterable[Variable] = None)

Convenience class for mapping a Variable object to anything else. This special map, however, supports accessing the image set both by the variable object instance itself _and_ its name.

data may be an iterable of (variable, value) pairs.

_variables

_map

property variables: set[Variable]

property varnames: dict[str, Variable]

property map: dict

__getitem__(key: str | Variable) → Any

__setitem__(variable: str | Variable, value: Any) → None

__delitem__(key: str | Variable) → None

__contains__(item: str | Variable) → bool

__iter__()

__len__()

__bool__()

__eq__(o: VariableMap)

__hash__()

__isub__(other)

__iadd__(other)

get(key: str | Variable, default: Any = None) → Any

keys() → collections.abc.Iterator[Variable]

values() → collections.abc.Iterator[Any]

items() → collections.abc.Iterator[tuple]

to_json() → dict[str, Any]

update(varmap: VariableMap) → VariableMap

copy(deep: bool = False) → VariableMap

classmethod from_json(variables: collections.abc.Iterable[Variable], d: dict[str, Any], typ=None, args=()) → VariableMap

__repr__()

class jpt.variables.VariableAssignment(data: collections.abc.Iterable[tuple] = None, variables: collections.abc.Iterable[Variable] = None)

Bases: VariableMap

Specialization of a VariableMap that maps a set of variables to values of the respective variables. This is an abstract base class that cannot be instantiated. There exist two specializations LabelAssignment and ValueAssignment that are supposed to be used instead.

data may be an iterable of (variable, value) pairs.

scalar2sets()

classmethod from_json(variables: collections.abc.Iterable[Variable], d: dict[str, Any], typ=None, args=()) → VariableMap

class jpt.variables.LabelAssignment(data: collections.abc.Iterable[tuple] = None, variables: collections.abc.Iterable[Variable] = None)

Bases: VariableAssignment

Maps a set of variables to values represented by their exterior representation, i.e. the perspective of a user.

data may be an iterable of (variable, value) pairs.

__setitem__(variable: Variable, value: set[int] | set[str] | jpt.base.intervals.NumberSet | numbers.Number | str) → None

value_assignment() → ValueAssignment

to_json() → dict[str, Any]

Convert this LabelAssignment to a json serializable dictionary. To achieve that sets are replaced with lists.

class jpt.variables.ValueAssignment(data: collections.abc.Iterable[tuple] = None, variables: collections.abc.Iterable[Variable] = None)

Bases: VariableAssignment

Maps a set of variables to values represented by their interior representation, i.e. the internal value representation used by JPTs.

data may be an iterable of (variable, value) pairs.

__setitem__(variable: Variable, value: set[int] | jpt.base.intervals.NumberSet | numbers.Number) → None

label_assignment() → LabelAssignment