Carl Weber – Lab 4

numpy_questions.py

@@ -37,10 +37,13 @@ def max_index(X):
         If the input is not a numpy array or
         if the shape is not 2D.
     """
-    i = 0
-    j = 0
+    if not isinstance(X, np.ndarray):
+        raise ValueError("Input must be a numpy array.")
+    if X.ndim != 2:
+        raise ValueError("Input must be a 2D array.")
 
-    # TODO
+    # argmax works on the flattened array, so convert back to (row, col)
+    i, j = np.unravel_index(np.argmax(X), X.shape)
 
     return i, j
 
@@ -62,6 +65,11 @@ def wallis_product(n_terms):
     pi : float
         The approximation of order `n_terms` of pi using the Wallis product.
     """
-    # XXX : The n_terms is an int that corresponds to the number of
-    # terms in the product. For example 10000.
-    return 0.
+    if n_terms == 0:
+        return 1.
+
+    k = np.arange(1, n_terms + 1)
+    terms = 4 * k**2 / (4 * k**2 - 1)
+
+    # the Wallis product converges to pi / 2
+    return 2 * np.prod(terms)

sklearn_questions.py

@@ -23,53 +23,82 @@
 from sklearn.base import BaseEstimator
 from sklearn.base import ClassifierMixin
 from sklearn.utils.validation import check_X_y
-from sklearn.utils.validation import check_array
 from sklearn.utils.validation import check_is_fitted
 from sklearn.utils.validation import validate_data
 from sklearn.utils.multiclass import check_classification_targets
 
 
 class OneNearestNeighbor(ClassifierMixin, BaseEstimator):
-    "OneNearestNeighbor classifier."
+    """OneNearestNeighbor classifier.
+
+    Predicts for each sample the label of the closest training sample,
+    using the Euclidean distance.
+    """
 
     def __init__(self):  # noqa: D107
         pass
 
     def fit(self, X, y):
-        """Write docstring.
-
-        And describe parameters
+        """Fit the classifier by storing the training data.
+
+        Parameters
+        ----------
+        X : ndarray of shape (n_samples, n_features)
+            The training samples.
+        y : ndarray of shape (n_samples,)
+            The training labels.
+
+        Returns
+        -------
+        self : OneNearestNeighbor
+            The fitted classifier.
         """
-        X, y = check_X_y(X, y)
+        X, y = validate_data(self, X, y)
         check_classification_targets(y)
         self.classes_ = np.unique(y)
-        self.n_features_in_ = X.shape[1]
 
-        # XXX fix
+        # 1-NN has no real training step, we just keep the data
+        self.X_train_ = X
+        self.y_train_ = y
         return self
 
     def predict(self, X):
-        """Write docstring.
+        """Predict the label of each sample.
 
-        And describe parameters
+        Parameters
+        ----------
+        X : ndarray of shape (n_samples, n_features)
+            The samples to classify.
+
+        Returns
+        -------
+        y_pred : ndarray of shape (n_samples,)
+            The label of the closest training sample for each sample.
         """
         check_is_fitted(self)
-        X = check_array(X)
-        y_pred = np.full(
-            shape=len(X), fill_value=self.classes_[0],
-            dtype=self.classes_.dtype
-        )
+        X = validate_data(self, X, reset=False)
 
-        # XXX fix
+        # distances between each sample and each training sample,
+        # shape (n_samples, n_train_samples)
+        dists = np.linalg.norm(X[:, np.newaxis] - self.X_train_, axis=2)
+        y_pred = self.y_train_[np.argmin(dists, axis=1)]
         return y_pred
 
     def score(self, X, y):
-        """Write docstring.
-
-        And describe parameters
+        """Compute the accuracy of the classifier.
+
+        Parameters
+        ----------
+        X : ndarray of shape (n_samples, n_features)
+            The test samples.
+        y : ndarray of shape (n_samples,)
+            The true labels.
+
+        Returns
+        -------
+        score : float
+            The fraction of correctly classified samples.
         """
         X, y = check_X_y(X, y)
         y_pred = self.predict(X)
-
-        # XXX fix
-        return y_pred.sum()
+        return np.mean(y_pred == y)