Added loss function and unit test for Poisson Loss

MLlib/loss_func.py

@@ -425,3 +425,62 @@ def loss(X, Y, W):
         y_pred = np.dot(X, W).T
         L = np.sum(np.true_divide((np.abs(Y - y_pred) * 100), Y)) / X.shape[0]
         return L
+
+
+class PoisonLoss():
+    """
+    Calculate Poisson Loss.
+    """
+
+    @staticmethod
+    def loss(X, Y, W):
+        """
+        Calculate Poisson Loss.
+
+        PARAMETERS
+        ==========
+
+        X: ndarray(dtype=float)
+            Input vector
+        Y: ndarray(dtype=float)
+            Output vector
+        W: ndarray(dtype=float)
+            Weights
+
+        RETURNS
+        =======
+
+        float or ndarray
+            array of Poisson losses or
+            float value of Poisson loss
+            (depending on the parameters)
+        """
+
+        y_pred = np.dot(X, W).T
+        return np.mean(y_pred - Y * np.log(y_pred), axis=-1)
+
+    @staticmethod
+    def derivative(X, Y, W):
+        """
+        Calculate derivative for Poisson Loss method.
+
+        PARAMETERS
+        ==========
+
+        X: ndarray(dtype=float)
+            Input vector
+        Y: ndarray(dtype=float)
+            Output vector
+        W: ndarray(dtype=float)
+            Weights
+
+        RETURNS
+        =======
+
+        ndarray
+            array of derivatives
+        """
+
+        M = X.shape[0]
+        y_pred = np.dot(X, W).T
+        return np.dot(1 - Y / y_pred, X) / M

MLlib/tests/test_loss_func.py

@@ -0,0 +1,26 @@
+from MLlib.loss_func import PoisonLoss
+import numpy as np
+
+
+def test_poisson_loss():
+    X = np.array([[0.32794909, 0.69075792],
+                  [0.9059869, 0.77822567],
+                  [0.77191135, 0.73130334],
+                  [0.78795424, 0.89521624],
+                  [0.96298287, 0.29203895],
+                  [0.17946441, 0.75993963],
+                  [0.05451229, 0.85694996],
+                  [0.2915702, 0.0805041],
+                  [0.57499626, 0.10640506],
+                  [0.95203463, 0.78138885],
+                  [0.51214417, 0.8854669],
+                  [0.94145457, 0.84719198]])
+    Y = np.array([[0.78936729, 0.8704933, 0.88695031, 0.98672232, 0.29796081,
+                   0.59689156, 0.88899604, 0.85520577, 0.18910425, 0.84729253,
+                   0.53142304, 0.7346871]])
+    W = np.array([[0.15736584],
+                  [0.3036601]])
+
+    poissonLoss = np.array([1.19509275])
+
+    assert np.isclose(PoisonLoss.loss(X, Y, W), poissonLoss).all()