Core classes

class pydimple.Graph(*args, **kwargs)

Bases: object

Computational graph. Consists of a set of Operators, Constants, and Distributions.

Parameters:

  • num_folds (int) – Number of folds to use in cross-fitting. Defaults to 5.

  • simplify (bool) – A boolean indicating whether to use sympy to try to symbolically evaluate when a random variable is known to be zero based on the conditioning set. Defaults to True.

class pydimple.Distribution(data, name=None, dtype=<class 'float'>)

Bases: Node

A data-generating distribution node in the computational graph. This holds estimates of any needed features of the distribution (conditional means, densities, etc.).

Parameters:

  • data (pd.DataFrame) – A DataFrame consisting of iid draws from the distribution.

  • name (str, optional) – Name of the distribution. Defaults to “Dist/”+count.

  • dtype (type, optional) – The type that the node holds, float, int, etc. Defaults to float.

class pydimple.Constant(value, name=None)

Bases: Node

Represents a constant value in the computational graph.

Parameters:

  • value (Any) – The value to assign to the constant. This value is immutable once set.

  • name (str, optional) – An optional name for the constant. If not provided, a unique name is generated.

Returns:

A new instance of Constant with the specified value and name.

Return type:

Constant

Raises:

ValueError – When attempting to modify the value of the constant after it is set.

class pydimple.Operator(*args, name='Operator')

Bases: Node

An operator node in the computational graph.

Parameters:

  • args (Node) – The input nodes to the operator.

  • name (str, optional) – The name of the operator. Defaults to “Operator”.

class pydimple.PointwiseFunction(func, unknown_vars=True, var_names=None, zero_if=None)

Bases: object

Represents a function that applies a mathematical operation pointwise.

Parameters:

  • func (callable) – A function.

  • unknown_vars (bool) – A boolean indicating whether the variables the function relies on are known. If this is true, then the var_names argument is ignored.

  • var_names (set or list) – A set or list of strings indicating the names of the variables the function relies on. Only used if unknown_vars is False. If a list is passed, it is converted to a set.

  • zero_if (str or sympy.Expr, optional) – Condition under which the function is a priori known to evaluate to zero.

Example:
>>> f = PointwiseFunction(lambda x: x ** 2, unknown_vars=False, var_names=["x"])
class pydimple.L2(func, P, unknown_vars=True, var_names=None, zero_if=None)

Bases: PointwiseFunction

Represents an L2 function, which is a subclass of PointwiseFunction with an associated Distribution object to represent the L2 space.

Parameters:

  • func (callable or PointwiseFunction) – A function or PointwiseFunction.

  • P (Distribution) – A Distribution object representing the L2 space.

  • unknown_vars (bool, optional) – A boolean indicating whether the variables the function relies on are known. If this is true, then the var_names argument is ignored. If func is a PointwiseFunction or L2, their unknown_vars values are used instead.

  • var_names (set or list, optional) – A set or list of strings indicating the names of the variables the function relies on. Only used if unknown_vars is False. If a list is passed, it is converted to a set. If func is a PointwiseFunction or L2, their var_names values are used instead.

  • zero_if (str or sympy.Expr, optional) – Condition under which the function is a priori known to evaluate to zero.

If func is already of type L2, the resulting L2 object will inherit its func and P attributes. The value of P can be updated by providing a new value in the P argument of the constructor, overriding the previous value.

Example:
>>> f = L2(lambda x: x ** 2, P=P)  # P is an existing Distribution object
>>> g = L2(f, P=Q)  # g inherits the func and var_names from f, but updates the value of P to an existing Distribution object Q