dataeval 0.66.0__py3-none-any.whl → 0.67.0__py3-none-any.whl

dataeval/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.66.0"
+__version__ = "0.67.0"
 
 from importlib.util import find_spec
 

dataeval/_internal/metrics/diversity.py

@@ -39,6 +39,13 @@ def diversity_shannon(
 
     Parameters
     ----------
+    data: NDArray
+        Array containing numerical values for metadata factors
+    names: list[str]
+        Names of metadata factors -- keys of the metadata dictionary
+    is_categorical: list[bool]
+        List of flags to identify whether variables are categorical (True) or
+        continuous (False)
     subset_mask: NDArray[np.bool_] | None
         Boolean mask of samples to bin (e.g. when computing per class).  True -> include in histogram counts
 
@@ -76,14 +83,20 @@ def diversity_simpson(
     Compute diversity for discrete/categorical variables and, through standard
     histogram binning, for continuous variables.
 
-    We define diversity as a normalized form of the inverse Simpson diversity
-    index.
+    We define diversity as the inverse Simpson diversity index linearly rescaled to the unit interval.
 
     diversity = 1 implies that samples are evenly distributed across a particular factor
-    diversity = 1/num_categories implies that all samples belong to one category/bin
+    diversity = 0 implies that all samples belong to one category/bin
 
     Parameters
     ----------
+    data: NDArray
+        Array containing numerical values for metadata factors
+    names: list[str]
+        Names of metadata factors -- keys of the metadata dictionary
+    is_categorical: list[bool]
+        List of flags to identify whether variables are categorical (True) or
+        continuous (False)
     subset_mask: NDArray[np.bool_] | None
         Boolean mask of samples to bin (e.g. when computing per class).  True -> include in histogram counts
 
@@ -91,10 +104,7 @@ def diversity_simpson(
     -----
     For continuous variables, histogram bins are chosen automatically.  See
         numpy.histogram for details.
-    The expression is undefined for q=1, but it approaches the Shannon entropy
-        in the limit.
-    If there is only one category, the diversity index takes a value of 1 =
-        1/N = 1/1.  Entropy will take a value of 0.
+    If there is only one category, the diversity index takes a value of 0.
 
     Returns
     -------
@@ -116,8 +126,8 @@ def diversity_simpson(
         # relative frequencies
         p_i = cnts / cnts.sum()
         # inverse Simpson index normalized by (number of bins)
-        ev_index[col] = 1 / np.sum(p_i**2) / num_bins[col]
-
+        s_0 = 1 / np.sum(p_i**2) / num_bins[col]
+        ev_index[col] = (s_0 * num_bins[col] - 1) / (num_bins[col] - 1)
     return ev_index
 
 
@@ -141,9 +151,8 @@ def diversity(
         List of class labels for each image
     metadata: List[Dict]
         List of metadata factors for each image
-    metric: Literal["shannon", "simpson"], default "simpson"
-        string variable indicating which diversity index should be used.
-        Permissible values include "simpson" and "shannon"
+    method: Literal["shannon", "simpson"], default "simpson"
+        Indicates which diversity index should be computed
 
     Notes
     -----
@@ -159,7 +168,7 @@ def diversity(
     Compute Simpson diversity index of metadata and class labels
 
     >>> diversity(class_labels, metadata, method="simpson").diversity_index
-    array([0.34482759, 0.34482759, 0.90909091])
+    array([0.18103448, 0.18103448, 0.88636364])
 
     Compute Shannon diversity index of metadata and class labels
 
@@ -189,7 +198,7 @@ def diversity_classwise(
     index.
 
     diversity = 1 implies that samples are evenly distributed across a particular factor
-    diversity = 1/num_categories implies that all samples belong to one category/bin
+    diversity = 0 implies that all samples belong to one category/bin
 
     Parameters
     ----------
@@ -197,12 +206,13 @@ def diversity_classwise(
         List of class labels for each image
     metadata: List[Dict]
         List of metadata factors for each image
+    method: Literal["shannon", "simpson"], default "simpson"
+        Indicates which diversity index should be computed
 
     Notes
     -----
     - For continuous variables, histogram bins are chosen automatically. See numpy.histogram for details.
-    - The expression is undefined for q=1, but it approaches the Shannon entropy in the limit.
-    - If there is only one category, the diversity index takes a value of 1 = 1/N = 1/1. Entropy will take a value of 0.
+    - If there is only one category, the diversity index takes a value of 0.
 
     Returns
     -------
@@ -214,8 +224,8 @@ def diversity_classwise(
     Compute classwise Simpson diversity index of metadata and class labels
 
     >>> diversity_classwise(class_labels, metadata, method="simpson").diversity_index
-    array([[0.33793103, 0.51578947],
-           [0.36      , 0.36      ]])
+    array([[0.17241379, 0.39473684],
+           [0.2       , 0.2       ]])
 
     Compute classwise Shannon diversity index of metadata and class labels
 

{dataeval-0.66.0.dist-info → dataeval-0.67.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: dataeval
-Version: 0.66.0
+Version: 0.67.0
 Summary: DataEval provides a simple interface to characterize image data and its impact on model performance across classification and object-detection tasks
 Home-page: https://dataeval.ai/
 License: MIT

{dataeval-0.66.0.dist-info → dataeval-0.67.0.dist-info}/RECORD

@@ -1,4 +1,4 @@
-dataeval/__init__.py,sha256=dshMbJco8lxfbbIg0DO5fSDsvgu4DKPGE5PzA7pwvPQ,590
+dataeval/__init__.py,sha256=RFLBI0XRq8nKsIoRacgjErr1RdQmoGI5gc8F9uAm-U8,590
 dataeval/_internal/detectors/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 dataeval/_internal/detectors/clusterer.py,sha256=hJwELUeAdZZ3OVLIfwalw2P7Zz13q2ZqrV6gx90s44E,20695
 dataeval/_internal/detectors/drift/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -24,7 +24,7 @@ dataeval/_internal/metrics/balance.py,sha256=g-YYFpq0qy2xq4iHjBKZDMjOn5R9Rit6sSb
 dataeval/_internal/metrics/ber.py,sha256=Onsi47AbT9rMvng-Pbu8LIrYRfLpI13En1FxkFoMKQs,4668
 dataeval/_internal/metrics/coverage.py,sha256=9ZvcNjItE9rEyA2UHPE1K9zpTbbib4xqk8WpPpDN8ok,4037
 dataeval/_internal/metrics/divergence.py,sha256=nmMUfr9FGnH798eb6xzEiMj4C42rQVthh5HeexiY6EE,4119
-dataeval/_internal/metrics/diversity.py,sha256=2xEkLnaRhPOvsd2DCTDT-dVvPPEZOH4PKm0vufrgBq4,8207
+dataeval/_internal/metrics/diversity.py,sha256=lkXnraNGCzkzamsBm2kG5eXxdRXHmkXxPWCKnGnPKOI,8620
 dataeval/_internal/metrics/parity.py,sha256=suv1Pf7gPj0_NxsS0_M6ewfUndsFJyEhbt5NPp6ktMI,15457
 dataeval/_internal/metrics/stats.py,sha256=Xbm7lLB0OZtsoxClMIrfULSqT8VymQiQmohJFtN7oz8,16332
 dataeval/_internal/metrics/uap.py,sha256=w-wvXXnX16kUq-weaZD2SrJi22LJ8EjOFbOhPxeGejI,2043
@@ -66,7 +66,7 @@ dataeval/torch/models/__init__.py,sha256=YnDnePYpRIKHyYn3F5qR1OObMSb-g0FGvI8X-uT
 dataeval/torch/trainer/__init__.py,sha256=Te-qElt8h-Zv8NN0r-VJOEdCPHTQ2yO3rd2MhRiZGZs,93
 dataeval/utils/__init__.py,sha256=ExQ1xj62MjcM9uIu1-g1P2fW0EPJpcIofnvxjQ908c4,172
 dataeval/workflows/__init__.py,sha256=gkU2B6yUiefexcYrBwqfZKNl8BvX8abUjfeNvVBXF4E,186
-dataeval-0.66.0.dist-info/LICENSE.txt,sha256=Kpzcfobf1HlqafF-EX6dQLw9TlJiaJzfgvLQFukyXYw,1060
-dataeval-0.66.0.dist-info/METADATA,sha256=P04dHyQOp4_6lg0IkoUEXTGJAPPpgRwf5ZAwdYpuatc,4217
-dataeval-0.66.0.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-dataeval-0.66.0.dist-info/RECORD,,
+dataeval-0.67.0.dist-info/LICENSE.txt,sha256=Kpzcfobf1HlqafF-EX6dQLw9TlJiaJzfgvLQFukyXYw,1060
+dataeval-0.67.0.dist-info/METADATA,sha256=KDbXi3-uY4kdbK4AuH4C9DXPLnkhfCdNAmqkHpDG1OI,4217
+dataeval-0.67.0.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+dataeval-0.67.0.dist-info/RECORD,,