Add kappa (#267)

AakashKumarNain · Squadrick · commit 977d96eacc27 · 2019-06-12T22:00:19.000+05:30
* Add Cohens Kappa Metric
diff --git a/tensorflow_addons/metrics/BUILD b/tensorflow_addons/metrics/BUILD
@@ -6,9 +6,23 @@ py_library(
     name = "metrics",
     srcs = [
         "__init__.py",
+        "cohens_kappa.py",
     ],
     srcs_version = "PY2AND3",
     deps = [
         "//tensorflow_addons/utils",
     ],
 )
+
+py_test(
+    name = "cohens_kappa_test",
+    size = "small",
+    srcs = [
+        "cohens_kappa_test.py",
+    ],
+    main = "cohens_kappa_test.py",
+    srcs_version = "PY2AND3",
+    deps = [
+        ":metrics",
+    ],
+)
diff --git a/tensorflow_addons/metrics/README.md b/tensorflow_addons/metrics/README.md
@@ -3,12 +3,12 @@
 ## Maintainers
 | Submodule  | Maintainers  | Contact Info   |
 |:---------- |:------------- |:--------------|
-|  |  |  |
+|  cohens_kappa| Aakash Nain   |  aakashnain@outlook.com|
 
 ## Contents
-| Submodule | Activation  | Reference                               |
+| Submodule | Metric  | Reference                               |
 |:----------------------- |:-------------------|:---------------|
-|  |  |            |
+| cohens_kappa| CohenKappa|[Cohen's Kappa](https://en.wikipedia.org/wiki/Cohen%27s_kappa)|
 
 
 ## Contribution Guidelines
diff --git a/tensorflow_addons/metrics/__init__.py b/tensorflow_addons/metrics/__init__.py
@@ -17,3 +17,5 @@
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
+
+from tensorflow_addons.metrics.cohens_kappa import CohenKappa
diff --git a/tensorflow_addons/metrics/cohens_kappa.py b/tensorflow_addons/metrics/cohens_kappa.py
@@ -0,0 +1,183 @@
+# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Implements Cohen's Kappa."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import tensorflow as tf
+import numpy as np
+import tensorflow.keras.backend as K
+from tensorflow.keras.metrics import Metric
+from tensorflow_addons.utils import keras_utils
+
+
+@keras_utils.register_keras_custom_object
+class CohenKappa(Metric):
+    """Computes Kappa score between two raters.
+
+    The score lies in the range [-1, 1]. A score of -1 represents
+    complete disagreement between two raters whereas a score of 1
+    represents complete agreement between the two raters.
+    A score of 0 means agreement by chance.
+
+    Note: As of now, this implementation considers all labels
+    while calculating the Cohen's Kappa score.
+
+    Usage:
+    ```python
+    actuals = np.array([4, 4, 3, 4, 2, 4, 1, 1], dtype=np.int32)
+    preds = np.array([4, 4, 3, 4, 4, 2, 1, 1], dtype=np.int32)
+
+    m = tf.keras.metrics.CohenKappa(num_classes=5)
+    m.update_state(actuals, preds, "quadratic")
+    print('Final result: ', m.result().numpy()) # Result: 0.68932
+    ```
+    Usage with tf.keras API:
+    ```python
+    model = keras.models.Model(inputs, outputs)
+    model.add_metric(tf.keras.metrics.CohenKappa(num_classes=5)(outputs))
+    model.compile('sgd', loss='mse')
+    ```
+
+    Args:
+      num_classes : Number of unique classes in your dataset
+      weightage   : Weighting to be considered for calculating
+                    kappa statistics. A valid value is one of
+                    [None, 'linear', 'quadratic']. Defaults to None.
+
+    Returns:
+      kappa_score : float
+        The kappa statistic, which is a number between -1 and 1. The maximum
+        value means complete agreement; zero or lower means chance agreement.
+
+    Raises:
+      ValueError: If the value passed for `weightage` is invalid
+        i.e. not any one of [None, 'linear', 'quadratic']
+    """
+
+    def __init__(self,
+                 num_classes,
+                 name='cohen_kappa',
+                 weightage=None,
+                 dtype=tf.float32):
+        super(CohenKappa, self).__init__(name=name, dtype=dtype)
+
+        if weightage not in (None, 'linear', 'quadratic'):
+            raise ValueError("Unknown kappa weighting type.")
+        else:
+            self.weightage = weightage
+
+        self.num_classes = num_classes
+        self.conf_mtx = self.add_weight(
+            'conf_mtx',
+            shape=(self.num_classes, self.num_classes),
+            initializer=tf.keras.initializers.zeros,
+            dtype=tf.int32)
+
+    def update_state(self, y_true, y_pred, sample_weight=None):
+        """Accumulates the confusion matrix condition statistics.
+
+        Args:
+          y_true : array, shape = [n_samples]
+                   Labels assigned by the first annotator.
+          y_pred : array, shape = [n_samples]
+                   Labels assigned by the second annotator. The kappa statistic
+                   is symmetric, so swapping ``y_true`` and ``y_pred`` doesn't
+                   change the value.
+          sample_weight(optional) : for weighting labels in confusion matrix
+                   Default is None. The dtype for weights should be the same
+                   as the dtype for confusion matrix. For more details,
+                   please check tf.math.confusion_matrix.
+
+
+        Returns:
+          Update op.
+        """
+        y_true = tf.cast(y_true, dtype=tf.int32)
+        y_pred = tf.cast(y_pred, dtype=tf.int32)
+
+        if y_true.shape != y_pred.shape:
+            raise ValueError(
+                "Number of samples in y_true and y_pred are different")
+
+        # compute the new values of the confusion matrix
+        new_conf_mtx = tf.math.confusion_matrix(
+            labels=y_true,
+            predictions=y_pred,
+            num_classes=self.num_classes,
+            weights=sample_weight)
+
+        # update the values in the original confusion matrix
+        return self.conf_mtx.assign_add(new_conf_mtx)
+
+    def result(self):
+        nb_ratings = tf.shape(self.conf_mtx)[0]
+        weight_mtx = tf.ones([nb_ratings, nb_ratings], dtype=tf.int32)
+
+        # 2. Create a weight matrix
+        if self.weightage is None:
+            diagonal = tf.zeros([nb_ratings], dtype=tf.int32)
+            weight_mtx = tf.linalg.set_diag(weight_mtx, diagonal=diagonal)
+            weight_mtx = tf.cast(weight_mtx, dtype=tf.float32)
+
+        else:
+            weight_mtx += tf.range(nb_ratings, dtype=tf.int32)
+            weight_mtx = tf.cast(weight_mtx, dtype=tf.float32)
+
+            if self.weightage == 'linear':
+                weight_mtx = tf.abs(weight_mtx - tf.transpose(weight_mtx))
+            else:
+                weight_mtx = tf.pow((weight_mtx - tf.transpose(weight_mtx)), 2)
+            weight_mtx = tf.cast(weight_mtx, dtype=tf.float32)
+
+        # 3. Get counts
+        actual_ratings_hist = tf.reduce_sum(self.conf_mtx, axis=1)
+        pred_ratings_hist = tf.reduce_sum(self.conf_mtx, axis=0)
+
+        # 4. Get the outer product
+        out_prod = pred_ratings_hist[..., None] * \
+                    actual_ratings_hist[None, ...]
+
+        # 5. Normalize the confusion matrix and outer product
+        conf_mtx = self.conf_mtx / tf.reduce_sum(self.conf_mtx)
+        out_prod = out_prod / tf.reduce_sum(out_prod)
+
+        conf_mtx = tf.cast(conf_mtx, dtype=tf.float32)
+        out_prod = tf.cast(out_prod, dtype=tf.float32)
+
+        # 6. Calculate Kappa score
+        numerator = tf.reduce_sum(conf_mtx * weight_mtx)
+        denominator = tf.reduce_sum(out_prod * weight_mtx)
+        kp = 1 - (numerator / denominator)
+        return kp
+
+    def get_config(self):
+        """Returns the serializable config of the metric."""
+
+        config = {
+            "num_classes": self.num_classes,
+            "weightage": self.weightage,
+        }
+        base_config = super(CohenKappa, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
+
+    def reset_states(self):
+        """Resets all of the metric state variables."""
+
+        for v in self.variables:
+            K.set_value(
+                v, np.zeros((self.num_classes, self.num_classes), np.int32))
diff --git a/tensorflow_addons/metrics/cohens_kappa_test.py b/tensorflow_addons/metrics/cohens_kappa_test.py
@@ -0,0 +1,135 @@
+# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Tests for Cohen's Kappa Metric."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import tensorflow as tf
+from tensorflow_addons.metrics import CohenKappa
+from tensorflow_addons.utils import test_utils
+
+
+@test_utils.run_all_in_graph_and_eager_modes
+class CohenKappaTest(tf.test.TestCase):
+    def test_config(self):
+        kp_obj = CohenKappa(name='cohen_kappa', num_classes=5)
+        self.assertEqual(kp_obj.name, 'cohen_kappa')
+        self.assertEqual(kp_obj.dtype, tf.float32)
+        self.assertEqual(kp_obj.num_classes, 5)
+
+        # Check save and restore config
+        kb_obj2 = CohenKappa.from_config(kp_obj.get_config())
+        self.assertEqual(kb_obj2.name, 'cohen_kappa')
+        self.assertEqual(kb_obj2.dtype, tf.float32)
+        self.assertEqual(kp_obj.num_classes, 5)
+
+    def initialize_vars(self):
+        kp_obj1 = CohenKappa(num_classes=5)
+        kp_obj2 = CohenKappa(num_classes=5, weightage='linear')
+        kp_obj3 = CohenKappa(num_classes=5, weightage='quadratic')
+
+        self.evaluate(tf.compat.v1.variables_initializer(kp_obj1.variables))
+        self.evaluate(tf.compat.v1.variables_initializer(kp_obj2.variables))
+        self.evaluate(tf.compat.v1.variables_initializer(kp_obj3.variables))
+        return kp_obj1, kp_obj2, kp_obj3
+
+    def update_obj_states(self, obj1, obj2, obj3, actuals, preds, weights):
+        update_op1 = obj1.update_state(actuals, preds, sample_weight=weights)
+        update_op2 = obj2.update_state(actuals, preds, sample_weight=weights)
+        update_op3 = obj3.update_state(actuals, preds, sample_weight=weights)
+
+        self.evaluate(update_op1)
+        self.evaluate(update_op2)
+        self.evaluate(update_op3)
+
+    def check_results(self, objs, values):
+        obj1, obj2, obj3 = objs
+        val1, val2, val3 = values
+
+        self.assertAllClose(val1, self.evaluate(obj1.result()), atol=1e-5)
+        self.assertAllClose(val2, self.evaluate(obj2.result()), atol=1e-5)
+        self.assertAllClose(val3, self.evaluate(obj3.result()), atol=1e-5)
+
+    def test_kappa_random_score(self):
+        actuals = [4, 4, 3, 4, 2, 4, 1, 1]
+        preds = [4, 4, 3, 4, 4, 2, 1, 1]
+        actuals = tf.constant(actuals, dtype=tf.int32)
+        preds = tf.constant(preds, dtype=tf.int32)
+
+        # Initialize
+        kp_obj1, kp_obj2, kp_obj3 = self.initialize_vars()
+
+        # Update
+        self.update_obj_states(kp_obj1, kp_obj2, kp_obj3, actuals, preds, None)
+
+        # Check results
+        self.check_results([kp_obj1, kp_obj2, kp_obj3],
+                           [0.61904761, 0.62790697, 0.68932038])
+
+    def test_kappa_perfect_score(self):
+        actuals = [4, 4, 3, 3, 2, 2, 1, 1]
+        preds = [4, 4, 3, 3, 2, 2, 1, 1]
+        actuals = tf.constant(actuals, dtype=tf.int32)
+        preds = tf.constant(preds, dtype=tf.int32)
+
+        # Initialize
+        kp_obj1, kp_obj2, kp_obj3 = self.initialize_vars()
+
+        # Update
+        self.update_obj_states(kp_obj1, kp_obj2, kp_obj3, actuals, preds, None)
+
+        # Check results
+        self.check_results([kp_obj1, kp_obj2, kp_obj3], [1.0, 1.0, 1.0])
+
+    def test_kappa_worse_than_random(self):
+        actuals = [4, 4, 3, 3, 2, 2, 1, 1]
+        preds = [1, 2, 4, 1, 3, 3, 4, 4]
+        actuals = tf.constant(actuals, dtype=tf.int32)
+        preds = tf.constant(preds, dtype=tf.int32)
+
+        # Initialize
+        kp_obj1, kp_obj2, kp_obj3 = self.initialize_vars()
+
+        # Update
+        self.update_obj_states(kp_obj1, kp_obj2, kp_obj3, actuals, preds, None)
+
+        # check results
+        self.check_results([kp_obj1, kp_obj2, kp_obj3],
+                           [-0.3333333, -0.52380952, -0.72727272])
+
+    def test_kappa_with_sample_weights(self):
+        actuals = [4, 4, 3, 3, 2, 2, 1, 1]
+        preds = [1, 2, 4, 1, 3, 3, 4, 4]
+        weights = [1, 1, 2, 5, 10, 2, 3, 3]
+        actuals = tf.constant(actuals, dtype=tf.int32)
+        preds = tf.constant(preds, dtype=tf.int32)
+        weights = tf.constant(weights, dtype=tf.int32)
+
+        # Initialize
+        kp_obj1, kp_obj2, kp_obj3 = self.initialize_vars()
+
+        # Update
+        self.update_obj_states(kp_obj1, kp_obj2, kp_obj3, actuals, preds,
+                               weights)
+
+        # check results
+        self.check_results([kp_obj1, kp_obj2, kp_obj3],
+                           [-0.25473321, -0.38992332, -0.60695344])
+
+
+if __name__ == '__main__':
+    tf.test.main()
