From 433cb958a4f32c64f78be79976702d0c48da9878 Mon Sep 17 00:00:00 2001
From: Nils Wisiol <mail@nils-wisiol.de>
Date: Wed, 12 Jan 2022 15:26:03 +0100
Subject: [PATCH] LR attack: allow setting validation set size

---
 pypuf/attack/lr2021.py | 14 ++++++--------
 1 file changed, 6 insertions(+), 8 deletions(-)

diff --git a/pypuf/attack/lr2021.py b/pypuf/attack/lr2021.py
index c6261a87..7c57b8b2 100644
--- a/pypuf/attack/lr2021.py
+++ b/pypuf/attack/lr2021.py
@@ -33,7 +33,7 @@ def on_epoch_end(self, epoch: int, logs: dict = None) -> None:
                 self.model.stop_training = True
 
     def __init__(self, crps: ChallengeResponseSet, seed: int, k: int, bs: int, lr: float, epochs: int,
-                 stop_validation_accuracy: float = .95) -> None:
+                 stop_validation_accuracy: float = .95, validation_set_size: float = .01) -> None:
         """
         Initialize an improved Logistic Regression attack using the given parameters.
 
@@ -56,6 +56,8 @@ def __init__(self, crps: ChallengeResponseSet, seed: int, k: int, bs: int, lr: f
         :param stop_validation_accuracy: Training is stopped when this validation accuracy is reached. Set to 1 to
             deactivate.
         :type stop_validation_accuracy: ``float``
+        :param validation_set_size: Proportion of CRPs to be used for validation, if <= 1, or else absolute number of
+            CRPs used to validation.
         """
         super().__init__(crps)
         self.crps = crps
@@ -67,6 +69,7 @@ def __init__(self, crps: ChallengeResponseSet, seed: int, k: int, bs: int, lr: f
         self.stop_validation_accuracy = stop_validation_accuracy
         self._history = None
         self._keras_model = None
+        self.validation_split = validation_set_size if validation_set_size <= 1 else validation_set_size / len(crps)
 
     @property
     def history(self) -> Optional[dict]:
@@ -147,7 +150,7 @@ def fit(self, verbose: bool = True) -> Simulation:
             features, labels,
             batch_size=self.bs,
             epochs=self.epochs,
-            validation_split=.01,
+            validation_split=self.validation_split,
             callbacks=[self.AccuracyStop(self.stop_validation_accuracy)],
             verbose=verbose,
         ).history
@@ -182,11 +185,6 @@ class BeliLR(LRAttack2021):
     def beli_output(output_delays: List[tf.Tensor]) -> List[tf.Tensor]:
         raise NotImplementedError
 
-    def __init__(self, crps: ChallengeResponseSet, seed: int, k: int, bs: int, lr: float, epochs: int,
-                 stop_validation_accuracy: float = .95, validation_set_size: int = 1000) -> None:
-        super().__init__(crps, seed, k, bs, lr, epochs, stop_validation_accuracy)
-        self.validation_set_size = validation_set_size
-
     def beli_model(self, input_tensor: tf.Tensor) -> List[tf.Tensor]:
         internal_delays = tf.keras.layers.Dense(
             units=1,
@@ -231,7 +229,7 @@ def fit(self, verbose: bool = True) -> Simulation:
             self.crps.responses,
             batch_size=self.bs,
             epochs=self.epochs,
-            validation_split=self.validation_set_size / N,  # validation set of 1000 CRPs
+            validation_split=self.validation_split,
             callbacks=[self.AccuracyStop(self.stop_validation_accuracy)],
             verbose=verbose,
         ).history