Allow AP preference to be any Nx computation, default to reduce_min (e…

…lixir-nx#198)

lib/scholar/cluster/affinity_propagation.ex

@@ -40,14 +40,16 @@ defmodule Scholar.Cluster.AffinityPropagation do
       """
     ],
     preference: [
-      type: :float,
+      type: {:or, [:float, :atom]},
+      default: :reduce_min,
       doc: """
-      Preferences for each point - points with larger values of preferences
-      are more likely to be chosen as exemplars. The number of clusters is
-      influenced by this option. If the preferences are not passed as arguments,
-      they will be set to the median of the input similarities (resulting in a
-      moderate number of clusters). For a smaller amount of clusters, this can
-      be set to the minimum value of the similarities.
+      How to compute the preferences for each point - points with larger values
+      of preferences are more likely to be chosen as exemplars. The number of clusters is
+      influenced by this option.
+
+      The preferences is either an atom, each is a `Nx` reduction function to
+      apply on the input similarities (such as `:reduce_min`, `:median`, `:mean`,
+      etc) or a float.
       """
     ],
     key: [
@@ -94,21 +96,21 @@ defmodule Scholar.Cluster.AffinityPropagation do
   ## Examples
 
       iex> key = Nx.Random.key(42)
-      iex> x = Nx.tensor([[12,5,78,2], [1,-5,7,32], [-1,3,6,1], [1,-2,5,2]])
+      iex> x = Nx.tensor([[12,5,78,2], [9,3,81,-2], [-1,3,6,1], [1,-2,5,2]])
       iex> Scholar.Cluster.AffinityPropagation.fit(x, key: key)
       %Scholar.Cluster.AffinityPropagation{
-        labels: Nx.tensor([0, 3, 3, 3]),
-        cluster_centers_indices: Nx.tensor([0, -1, -1, 3]),
+        labels: Nx.tensor([0, 0, 2, 2]),
+        cluster_centers_indices: Nx.tensor([0, -1, 2, -1]),
         cluster_centers: Nx.tensor(
           [
             [12.0, 5.0, 78.0, 2.0],
             [:infinity, :infinity, :infinity, :infinity],
-            [:infinity, :infinity, :infinity, :infinity],
-            [1.0, -2.0, 5.0, 2.0]
+            [-1.0, 3.0, 6.0, 1.0],
+            [:infinity, :infinity, :infinity, :infinity]
           ]
         ),
         num_clusters: Nx.tensor(2, type: :u64),
-        iterations: Nx.tensor(18, type: :s64)
+        iterations: Nx.tensor(22, type: :s64)
       }
   """
   deftransform fit(data, opts \\ []) do
@@ -233,14 +235,15 @@ defmodule Scholar.Cluster.AffinityPropagation do
   defnp initialize_similarity(data, opts \\ []) do
     n = Nx.axis_size(data, 0)
     dist = -Scholar.Metrics.Distance.pairwise_squared_euclidean(data)
+    preference = initialize_preference(dist, opts[:preference])
+    Nx.put_diagonal(dist, Nx.broadcast(preference, {n}))
+  end
 
-    fill_in =
-      case opts[:preference] do
-        nil -> Nx.broadcast(Nx.median(dist), {n})
-        preference -> Nx.broadcast(preference, {n})
-      end
-
-    Nx.put_diagonal(dist, fill_in)
+  deftransformp initialize_preference(dist, preference) do
+    cond do
+      is_atom(preference) -> apply(Nx, preference, [dist])
+      is_float(preference) -> preference
+    end
   end
 
   @doc """
@@ -251,20 +254,20 @@ defmodule Scholar.Cluster.AffinityPropagation do
   ## Examples
 
       iex> key = Nx.Random.key(42)
-      iex> x = Nx.tensor([[12,5,78,2], [1,-5,7,32], [-1,3,6,1], [1,-2,5,2]])
+      iex> x = Nx.tensor([[12,5,78,2], [9,3,81,-2], [-1,3,6,1], [1,-2,5,2]])
       iex> model = Scholar.Cluster.AffinityPropagation.fit(x, key: key)
       iex> Scholar.Cluster.AffinityPropagation.prune(model)
       %Scholar.Cluster.AffinityPropagation{
-        labels: Nx.tensor([0, 1, 1, 1]),
-        cluster_centers_indices: Nx.tensor([0, 3]),
+        labels: Nx.tensor([0, 0, 1, 1]),
+        cluster_centers_indices: Nx.tensor([0, 2]),
         cluster_centers: Nx.tensor(
           [
             [12.0, 5.0, 78.0, 2.0],
-            [1.0, -2.0, 5.0, 2.0]
+            [-1.0, 3.0, 6.0, 1.0]
           ]
         ),
         num_clusters: Nx.tensor(2, type: :u64),
-        iterations: Nx.tensor(18, type: :s64)
+        iterations: Nx.tensor(22, type: :s64)
       }
   """
   def prune(
@@ -302,13 +305,13 @@ defmodule Scholar.Cluster.AffinityPropagation do
   ## Examples
 
       iex> key = Nx.Random.key(42)
-      iex> x = Nx.tensor([[12,5,78,2], [1,5,7,32], [1,3,6,1], [1,2,5,2]])
+      iex> x = Nx.tensor([[12,5,78,2], [9,3,81,-2], [-1,3,6,1], [1,-2,5,2]])
       iex> model = Scholar.Cluster.AffinityPropagation.fit(x, key: key)
       iex> model = Scholar.Cluster.AffinityPropagation.prune(model)
-      iex> Scholar.Cluster.AffinityPropagation.predict(model, Nx.tensor([[1,6,2,6], [8,3,8,2]]))
+      iex> Scholar.Cluster.AffinityPropagation.predict(model, Nx.tensor([[10,3,50,6], [8,3,8,2]]))
       #Nx.Tensor<
         s64[2]
-        [1, 1]
+        [0, 1]
       >
   """
   defn predict(%__MODULE__{cluster_centers: cluster_centers} = _model, x) do

test/scholar/cluster/affinity_propagation_test.exs

@@ -68,7 +68,7 @@ defmodule Scholar.Cluster.AffinityPropagationTest do
   end
 
   test "fit and compute_values" do
-    model = AffinityPropagation.fit(x(), key: key())
+    model = AffinityPropagation.fit(x(), key: key(), preference: :median)
 
     model = AffinityPropagation.prune(model)
 
@@ -94,14 +94,14 @@ defmodule Scholar.Cluster.AffinityPropagationTest do
   end
 
   test "predict with pruning" do
-    model = AffinityPropagation.fit(x(), key: key())
+    model = AffinityPropagation.fit(x(), key: key(), preference: :median)
     model = AffinityPropagation.prune(model)
     preds = AffinityPropagation.predict(model, x_test())
     assert preds == Nx.tensor([0, 2, 0, 5, 5, 5, 2, 2, 5, 2])
   end
 
   test "predict without pruning" do
-    model = AffinityPropagation.fit(x(), key: key())
+    model = AffinityPropagation.fit(x(), key: key(), preference: :median)
     preds = AffinityPropagation.predict(model, x_test())
     assert preds == Nx.tensor([2, 9, 2, 34, 34, 34, 9, 9, 34, 9])
   end