Add Doctest to back_propogation_neural_network and huffman.py

Author: dhruvi003

data_compression/huffman.py

@@ -39,6 +39,13 @@ def build_tree(letters: list[Letter]) -> Letter | TreeNode:
     """
     Run through the list of Letters and build the min heap
     for the Huffman Tree.
+
+    >>> letters = [Letter('a', 5), Letter('b', 9), Letter('c', 12), Letter('d', 13)]
+    >>> root = build_tree(letters)
+    >>> isinstance(root, TreeNode)
+    True
+    >>> root.freq
+    39
     """
     response: list[Letter | TreeNode] = list(letters)
     while len(response) > 1:
@@ -55,6 +62,16 @@ def traverse_tree(root: Letter | TreeNode, bitstring: str) -> list[Letter]:
     """
     Recursively traverse the Huffman Tree to set each
     Letter's bitstring dictionary, and return the list of Letters
+
+     >>> letters = [Letter('a', 2), Letter('b', 3), Letter('c', 4)]
+    >>> root = build_tree(letters)
+    >>> result = traverse_tree(root, "")
+    >>> sorted([l.letter for l in result])
+    ['a', 'b', 'c']
+    >>> all(l.bitstring[l.letter] for l in result)
+    True
+    >>> sum(l.freq for l in result)
+    9
     """
     if isinstance(root, Letter):
         root.bitstring[root.letter] = bitstring

neural_network/back_propagation_neural_network.py

@@ -23,6 +23,15 @@
 
 
 def sigmoid(x: np.ndarray) -> np.ndarray:
+    """
+    Compute the sigmoid activation function
+
+    >>> import numpy as np
+    >>> np.allclose(sigmoid(np.array([0])), np.array([0.5]))
+    True
+    >>> np.allclose(sigmoid(np.array([-1, 0, 1])), np.array([0.26894142, 0.5, 0.73105858]))
+    True
+    """
     return 1 / (1 + np.exp(-x))
 
 
@@ -158,6 +167,19 @@ def train(self, xdata, ydata, train_round, accuracy):
         return None
 
     def cal_loss(self, ydata, ydata_):
+        """
+        Calculate Mean Squared Error (MSE) loss and its gradient.
+
+        >>> import numpy as np
+        >>> bp = BPNN()
+        >>> y_true = np.asmatrix([[1.0], [0.5]])
+        >>> y_pred = np.asmatrix([[0.8], [0.3]])
+        >>> loss, grad = bp.cal_loss(y_true, y_pred)
+        >>> float(round(loss, 2))
+        0.08
+        >>> np.allclose(grad, np.array([[-0.4], [-0.4]]))
+        True
+        """
         self.loss = np.sum(np.power((ydata - ydata_), 2))
         self.loss_gradient = 2 * (ydata_ - ydata)
         # vector (shape is the same as _ydata.shape)