y must be of rank 2

krstopro · May 16, 2024 · 27e4f77 · 27e4f77
1 parent 9e8f35a
commit 27e4f77
Show file tree

Hide file tree

Showing 2 changed files with 18 additions and 22 deletions.
diff --git a/lib/scholar/neighbors/knn_regressor.ex b/lib/scholar/neighbors/knn_regressor.ex
@@ -55,43 +55,43 @@ defmodule Scholar.Neighbors.KNNRegressor do
   ## Examples
 
       iex> x = Nx.tensor([[1, 2], [2, 3], [3, 4], [4, 5], [5, 6]])
-      iex> y = Nx.tensor([1, 2, 3, 4, 5])
+      iex> y = Nx.tensor([[1], [2], [3], [4], [5]])
       iex> model = Scholar.Neighbors.KNNRegressor.fit(x, y, num_neighbors: 3)
       iex> model.algorithm
       Scholar.Neighbors.BruteKNN.fit(x, num_neighbors: 3)
       iex> model.labels
-      Nx.tensor([1, 2, 3, 4, 5])
+      Nx.tensor([[1], [2], [3], [4], [5]])
 
       iex> x = Nx.tensor([[1, 2], [2, 3], [3, 4], [4, 5], [5, 6]])
-      iex> y = Nx.tensor([1, 2, 3, 4, 5])
+      iex> y = Nx.tensor([[1], [2], [3], [4], [5]])
       iex> model = Scholar.Neighbors.KNNRegressor.fit(x, y, algorithm: :kd_tree, num_neighbors: 3, metric: {:minkowski, 1})
       iex> model.algorithm
       Scholar.Neighbors.KDTree.fit(x, num_neighbors: 3, metric: {:minkowski, 1})
       iex> model.labels
-      Nx.tensor([1, 2, 3, 4, 5])
+      Nx.tensor([[1], [2], [3], [4], [5]])
 
       iex> x = Nx.tensor([[1, 2], [2, 3], [3, 4], [4, 5], [5, 6]])
-      iex> y = Nx.tensor([1, 2, 3, 4, 5])
+      iex> y = Nx.tensor([[1], [2], [3], [4], [5]])
       iex> key = Nx.Random.key(12)
       iex> model = Scholar.Neighbors.KNNRegressor.fit(x, y, algorithm: :random_projection_forest, num_neighbors: 2, num_trees: 4, key: key)
       iex> model.algorithm
       Scholar.Neighbors.RandomProjectionForest.fit(x, num_neighbors: 2, num_trees: 4, key: key)
       iex> model.labels
-      Nx.tensor([1, 2, 3, 4, 5])
+      Nx.tensor([[1], [2], [3], [4], [5]])
   """
   deftransform fit(x, y, opts) do
     if Nx.rank(x) != 2 do
       raise ArgumentError,
             """
-            expected x to have shape {num_samples, num_features}, \
+            expected x to have shape {num_samples, num_features_in}, \
             got tensor with shape: #{inspect(Nx.shape(x))}
             """
     end
 
-    if Nx.rank(y) not in [1, 2] do
+    if Nx.rank(y) != 2 do
       raise ArgumentError,
             """
-            expected y to have rank 1 or 2, \
+            expected y to have shape {num_samples, num_features_out}, \
             got tensor with shape: #{inspect(Nx.shape(y))}
             """
     end
@@ -137,11 +137,11 @@ defmodule Scholar.Neighbors.KNNRegressor do
   ## Examples
 
       iex> x_train = Nx.tensor([[1, 2], [2, 3], [3, 4], [4, 5], [5, 6]])
-      iex> y_train = Nx.tensor([1, 2, 3, 4, 5])
+      iex> y_train = Nx.tensor([[1], [2], [3], [4], [5]])
       iex> model = Scholar.Neighbors.KNNRegressor.fit(x_train, y_train, num_neighbors: 3)
       iex> x = Nx.tensor([[1, 3], [4, 2], [3, 6]])
       iex> Scholar.Neighbors.KNNRegressor.predict(model, x)
-      Nx.tensor([2.0, 2.0, 4.0])
+      Nx.tensor([[2.0], [2.0], [4.0]])
   """
   defn predict(model, x) do
     {neighbors, distances} = compute_knn(model.algorithm, x)
@@ -152,14 +152,10 @@ defmodule Scholar.Neighbors.KNNRegressor do
         Nx.mean(neighbor_labels, axes: [1])
 
       :distance ->
-        weights = Scholar.Neighbors.Utils.check_weights(distances)
-
         weights =
-          if Nx.rank(model.labels) == 2 do
-            weights |> Nx.new_axis(2) |> Nx.broadcast(neighbor_labels)
-          else
-            weights
-          end
+          Scholar.Neighbors.Utils.check_weights(distances)
+          |> Nx.new_axis(2)
+          |> Nx.broadcast(neighbor_labels)
 
         Nx.weighted_mean(neighbor_labels, weights, axes: [1])
     end

diff --git a/test/scholar/neighbors/knn_regressor_test.exs b/test/scholar/neighbors/knn_regressor_test.exs
@@ -19,7 +19,7 @@ defmodule Scholar.Neighbors.KNNRegressorTest do
   end
 
   defp y_train do
-    Nx.tensor([0, 1, 1, 1, 1, 1, 1, 1, 0, 0])
+    Nx.tensor([[0], [1], [1], [1], [1], [1], [1], [1], [0], [0]])
   end
 
   defp x do
@@ -70,13 +70,13 @@ defmodule Scholar.Neighbors.KNNRegressorTest do
     test "predict with default parameters" do
       model = KNNRegressor.fit(x_train(), y_train(), num_neighbors: 3)
       y_pred = KNNRegressor.predict(model, x())
-      assert_all_close(y_pred, Nx.tensor([0.66666667, 0.66666667, 0.33333333, 1.0]))
+      assert_all_close(y_pred, Nx.tensor([[0.66666667], [0.66666667], [0.33333333], [1.0]]))
     end
 
     test "predict with weights set to :distance" do
       model = KNNRegressor.fit(x_train(), y_train(), num_neighbors: 3, weights: :distance)
       y_pred = KNNRegressor.predict(model, x())
-      assert_all_close(y_pred, Nx.tensor([0.59648849, 0.68282796, 0.2716506, 1.0]))
+      assert_all_close(y_pred, Nx.tensor([[0.59648849], [0.68282796], [0.2716506], [1.0]]))
     end
 
     test "predict with cosine metric and weights set to :distance" do
@@ -88,7 +88,7 @@ defmodule Scholar.Neighbors.KNNRegressorTest do
         )
 
       y_pred = KNNRegressor.predict(model, x())
-      assert_all_close(y_pred, Nx.tensor([0.5736568, 0.427104, 0.33561941, 1.0]))
+      assert_all_close(y_pred, Nx.tensor([[0.5736568], [0.427104], [0.33561941], [1.0]]))
     end
 
     test "predict with 2D labels" do