PyEstimator Class Reference

Class to invoke a pretrained python estimator that follows the sklearn interface. More...

#include <PyEstimator.h>

Collaboration diagram for PyEstimator:

Public Member Functions
	PyEstimator (const std::string &pickleFileName)
	Construct the Estimator. More...
double	predict (const std::vector< double > &inputVariables)
	Call the predict method of the estimator.
double	predict (const std::vector< NamedFloatTuple * > &floatTuples)
	Call the predict method of the estimator.
double	predict (boost::python::object &array)
	Call the predict method of the estimator.

Private Member Functions
void	unpickleEstimator (const std::string &pickleFileName)
	Load the estimator object from the pickled file.
void	expand (size_t nVars)
	Reserves space for at least n variable in the input array.

Private Attributes
std::string	m_pickleFileName
	File name of the pickle file that contains the trained estimator.
boost::python::object	m_estimator
	Retrained python estimator object.
boost::python::object	m_predict
	Python bound prediction method - cached to avoid repeated lookup.
boost::python::object	m_array
	Array to be served to the estimator.
size_t	m_nCurrent = 0
	Cache for the current length of the input array.
bool	m_is_binary_classification = false
	Internal flag to keep track whether a binary classification with predict_proba is evaluated.

Detailed Description

Class to invoke a pretrained python estimator that follows the sklearn interface.

Definition at line 23 of file PyEstimator.h.

Constructor & Destructor Documentation

PyEstimator()

PyEstimator ( const std::string &  pickleFileName )

explicit

Construct the Estimator.

Parameters

pickleFileName

Name of the file that contains the pickled estimator object eg from sklearn

Definition at line 22 of file PyEstimator.cc.

  : m_pickleFileName(pickleFileName)
{
  try {
    // Construct an array with one entry
    // Expand it once the number of variables is known.
    boost::python::object numpy = boost::python::import("numpy");
    boost::python::list initValues;
    initValues.append(0.0);
    m_array = numpy.attr("array")(initValues);
    m_nCurrent = boost::python::len(m_array);
    // boost::python::object array = boost::python::import("array");
    // m_array = array.attr("array")("d");
    unpickleEstimator(pickleFileName);
  } catch (const boost::python::error_already_set&) {
    PyErr_Print();
    B2ERROR("Could not construct PyEstimator from " << pickleFileName);
  }
}

---

The documentation for this class was generated from the following files:

• tracking/trackFindingCDC/mva/include/PyEstimator.h
• tracking/trackFindingCDC/mva/src/PyEstimator.cc