Deprecate shuffle prop in favor of PowerSampling

baal/active/heuristics/heuristics.py

@@ -12,6 +12,13 @@
 
 from baal.utils.array_utils import to_prob
 
+DEPRECATED = "DEPRECATED"
+SHUFFLE_PROP_DEPRECATION_NOTICE = """
+`shuffle_prop` is deprecated and will be removed in Baal 1.9.0.
+For better batch uncertainty estimation, use `baal.active.heuristics.stochastics.PowerSampling`.
+See `https://baal.readthedocs.io/en/latest/user_guide/heuristics/` for more details.
+"""
+
 available_reductions = {
     "max": lambda x: np.max(x, axis=tuple(range(1, x.ndim))),
     "min": lambda x: np.min(x, axis=tuple(range(1, x.ndim))),
@@ -139,7 +146,11 @@ class AbstractHeuristic:
         reduction (Union[str, Callable]): Reduction used after computing the score.
     """
 
-    def __init__(self, shuffle_prop=0.0, reverse=False, reduction="none"):
+    def __init__(self, shuffle_prop=DEPRECATED, reverse=False, reduction="none"):
+        if shuffle_prop != DEPRECATED and shuffle_prop < 1.0:
+            warnings.warn(SHUFFLE_PROP_DEPRECATION_NOTICE, DeprecationWarning)
+        else:
+            shuffle_prop = 0.0
         self.shuffle_prop = shuffle_prop
         self.reversed = reverse
         assert reduction in available_reductions or callable(reduction)
@@ -272,7 +283,7 @@ class BALD(AbstractHeuristic):
         https://arxiv.org/abs/1703.02910
     """
 
-    def __init__(self, shuffle_prop=0.0, reduction="none"):
+    def __init__(self, shuffle_prop=DEPRECATED, reduction="none"):
         super().__init__(shuffle_prop=shuffle_prop, reverse=True, reduction=reduction)
 
     @require_single_item
@@ -324,7 +335,7 @@ class BatchBALD(BALD):
         Not tested on 4+ dims.
     """
 
-    def __init__(self, num_samples, num_draw=500, shuffle_prop=0.0, reduction="none"):
+    def __init__(self, num_samples, num_draw=500, shuffle_prop=DEPRECATED, reduction="none"):
         self.epsilon = 1e-5
         self.num_samples = num_samples
         self.num_draw = num_draw
@@ -508,7 +519,7 @@ class Variance(AbstractHeuristic):
         reduction (Union[str, callable]): function that aggregates the results (default: `mean`).
     """
 
-    def __init__(self, shuffle_prop=0.0, reduction="mean"):
+    def __init__(self, shuffle_prop=DEPRECATED, reduction="mean"):
         _help = "Need to reduce the output from [n_sample, n_class] to [n_sample]"
         assert reduction != "none", _help
         super().__init__(shuffle_prop=shuffle_prop, reverse=True, reduction=reduction)
@@ -529,7 +540,7 @@ class Entropy(AbstractHeuristic):
         reduction (Union[str, callable]): function that aggregates the results (default: `none`).
     """
 
-    def __init__(self, shuffle_prop=0.0, reduction="none"):
+    def __init__(self, shuffle_prop=DEPRECATED, reduction="none"):
         super().__init__(shuffle_prop=shuffle_prop, reverse=True, reduction=reduction)
 
     @require_single_item
@@ -551,7 +562,7 @@ class Margin(AbstractHeuristic):
             (default: `none`).
     """
 
-    def __init__(self, shuffle_prop=0.0, reduction="none"):
+    def __init__(self, shuffle_prop=DEPRECATED, reduction="none"):
         super().__init__(shuffle_prop=shuffle_prop, reverse=False, reduction=reduction)
 
     @require_single_item
@@ -571,7 +582,7 @@ class Certainty(AbstractHeuristic):
         reduction (Union[str, callable]): function that aggregates the results.
     """
 
-    def __init__(self, shuffle_prop=0.0, reduction="none"):
+    def __init__(self, shuffle_prop=DEPRECATED, reduction="none"):
         super().__init__(shuffle_prop=shuffle_prop, reverse=False, reduction=reduction)
 
     @require_single_item
@@ -588,7 +599,7 @@ class Precomputed(AbstractHeuristic):
         reverse (Bool): Sort from lowest to highest if False.
     """
 
-    def __init__(self, shuffle_prop=0.0, reverse=False):
+    def __init__(self, shuffle_prop=DEPRECATED, reverse=False):
         super().__init__(shuffle_prop, reverse=reverse)
 
     def compute_score(self, predictions):
@@ -604,7 +615,7 @@ class Random(Precomputed):
         seed (Optional[int]): If provided, will seed the random generator.
     """
 
-    def __init__(self, shuffle_prop=0.0, reduction="none", seed=None):
+    def __init__(self, shuffle_prop=DEPRECATED, reduction="none", seed=None):
         super().__init__(1.0, False)
         if seed is not None:
             self.rng = np.random.RandomState(seed)
@@ -643,7 +654,7 @@ class CombineHeuristics(AbstractHeuristic):
 
     """
 
-    def __init__(self, heuristics: List, weights: List, reduction="mean", shuffle_prop=0.0):
+    def __init__(self, heuristics: List, weights: List, reduction="mean", shuffle_prop=DEPRECATED):
         super(CombineHeuristics, self).__init__(reduction=reduction, shuffle_prop=shuffle_prop)
         self.composed_heuristic = heuristics
         self.weights = weights