Implementation of the threshouldout method

thresholdout.py

# Copyright (c) 2017 Mario Tambos

# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:

# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.

# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.

def thresholdout(scorer: Callable,
                 y_train: np.ndarray, y_train_pred: np.ndarray,
                 y_test: np.ndarray, y_test_pred: np.ndarray,
                 T: float=-1, tau: float=-1):
  """
  Based on:
  The reusable holdout: Preserving validity in adaptive data analysis
  BY CYNTHIA DWORK, VITALY FELDMAN, MORITZ HARDT, TONIANN PITASSI, OMER REINGOLD, AARON ROTH
  SCIENCE07 AUG 2015 : 636-638
  """
    if T <= 0:
        T = 1/len(y_train)
    if tau <= 0:
        tau = 4/len(y_train)
    T_hat = T + np.random.laplace(scale=tau)
    xi = np.random.laplace(scale=tau)
    eta = np.random.laplace(scale=tau)
    ESh = scorer(y_test, y_test_pred)
    ESt = scorer(y_train, y_train_pred)
    if np.abs(ESh - ESt) > T_hat + eta:
        return ESh + xi
    else:
        return ESt