API Documentation

This is the API documentation of the public API of multilstsq. These are the most commonly used functions when using multilstsq.

Least squares

class multilstsq.MultiLstSq(problem_dimensions, n_parameters, internal_dtype=<class 'float'>)

Multiple least squares problems.

y = Xβ + ε

This is a class which can exist in three modes:

  • MODE_REGRESSION: add data to the least square matrices to obtain the mean
  • MODE_VARIANCE: compute the variance
  • MODE_READONLY: the object is frozen.

The mode can only be switched “forward” (it is not possible to move from MODE_VARIANCE to MODE_REGRESSION for example.)

__init__(problem_dimensions, n_parameters, internal_dtype=<class 'float'>)

Create a MultiLstSq object, which is originally in MODE_REGRESSION.

Parameters:
  • problem_dimensions – Tuple, size of the problem array, it can be () (0-dimensional), for a single least square problem, or for example (800, 600) for 800x600 times the regression problem
  • n_parameters – Number of parameters of the least squares problem.
  • internal_dtype – numpy.dtype data type of the matrices
__weakref__

list of weak references to the object (if defined)

add_data(X, y, w=None)

Add data to the object (depending on the mode, for either mean or variance computation)

beta

The linear coefficients that minimize the least squares criterion.

evaluate_at(X)

Evaluate Xβ

n_observations

Number of observations n.

n_parameters

Number of parameters

rss

Residual sum of squares

sigma_2

Scaling parameter for the covariance matrix.

switch_to_read_only()

Switch to read-only mode.

switch_to_variance()

Switch to variance computation mode.

variance

Returns the variance/covariance matrix.

Regression

class multilstsq.MultiRegression(problem_dimensions, model_str, internal_dtype=<class 'float'>)
__init__(problem_dimensions, model_str, internal_dtype=<class 'float'>)

Create a MultiLstSq object, which is originally in MODE_REGRESSION.

Parameters:
  • problem_dimensions – Tuple, size of the problem array, it can be () (0-dimensional), for a single least square problem, or for example (800, 600) for 800x600 times the regression problem
  • n_parameters – Number of parameters of the least squares problem.
  • internal_dtype – numpy.dtype data type of the matrices
add_data(X, y, w=None)

Add data to the object (depending on the mode, for either mean or variance computation)

Expression evaluator

class multilstsq.ExprEvaluator(expr, constants=None, enable_caller_modules=True)
__call__(*args, **kwargs)

Substitute the arguments in the expression, and then evaluate it and return the resulting Python object.

The positional arguments are zip()’ed with the variable_list of multilstsq.ExprEvaluator.enable_call(), which the named arguments are directly substituted.

Returns:

The Python object of the evaluated expression.
Raises:
__init__(expr, constants=None, enable_caller_modules=True)

Initialize an ExprEvaluator object

Parameters:
  • expr – Expression string, should be a valid Python 3 expression
  • constant – Dictionary of constants {'constantname': value}.
  • enable_caller_modules – Boolean, if True modules of the first frame in the stack outside of multilstsq will be included in the context of the expression
__repr__()

Return repr(self).

__str__()
Returns:a string corresponding to the expression
__weakref__

list of weak references to the object (if defined)

constants
Returns:Set of constants names (values which have a defined substitution) in the current expression
enable_call(variable_list)

Enables calling the expression, as a simplification for calling multilstsq.ExprEvaluator.substitute() followed by multilstsq.ExprEvaluator.eval().

Parameters:variable_list – List of variables names which correspond to the arguments which will be used in multilstsq.ExprEvaluator.__call__()
eval()
Returns:The python object which results of the expression
Raises:ValueError – if at least one variable is not defined
reduce()
Returns:A copy of the current expression, where all known part of the syntax tree are simplified.

For example, reducing a*(b+c) with known b and c, will result in a*__ExprEvaluator_0, where __ExprEvaluator_0 is defined as a constant.

substitute(expressions=None, constants=None)

Substitute expressions or constants in the current expression.

Parameters:
  • expressions – dictionary of expression to substitute. The keys can be of str, ast.AST or multilstsq.ExprEvaluator.
  • constants – dictionary of constants to substitute {'constantname': value}
Returns:

New multilstsq.ExprEvaluator object with the substitution done.
variables
Returns:Set of variables names (values which don’t have a defined substitution) in the current expression