| |
| |
|
|
|
|
| class ThresholdOperation(): |
| """Represents the threshold operations that are used in postprocessing approaches. |
| |
| Threshold operations simply indicate a threshold and an operator, thereby defining a function. |
| The function can be evaluated at arbitrary points (usually the scores returned from |
| unconstrained predictors) to return a bool value. |
| |
| :param operator: the threshold operator, can be either '>' or '<' |
| :type operator: str |
| :param threshold: the threshold, can be numpy.inf or -numpy.inf |
| :type threshold: float |
| """ |
|
|
| def __init__(self, operator, threshold): |
| if operator not in ['>', '<']: |
| raise ValueError("Unrecognized operator: " + operator) |
| self._operator = operator |
| self._threshold = threshold |
|
|
| @property |
| def threshold(self): |
| """Return the stored threshold.""" |
| return self._threshold |
|
|
| @property |
| def operator(self): |
| """Return the stored threshold operator.""" |
| return self._operator |
|
|
| def get_predictor_from_operation(self): |
| """Encode the threshold rule `Y_hat > t` or `Y_hat < t`. |
| |
| :return: a function that takes a single argument to evaluate it against the threshold rule |
| :rtype: lambda |
| """ |
| if self._operator == '>': |
| return lambda x: x > self._threshold |
| elif self._operator == '<': |
| return lambda x: x < self._threshold |
| else: |
| raise ValueError("Unrecognized operator: " + self._operator) |
|
|
| def __repr__(self): |
| return "[{}{}]".format(self._operator, self._threshold) |
|
|
