Update PCA doctests for Nx 0.13's eigh fix

Verified against sklearn on the iris dataset (fit, incremental_fit,
partial_fit, transform) and against a numpy SVD reference (fit_transform,
n=6). The n=6 case is the clearest evidence this is a real precision fix,
not just reformatting: the old second-component value was off by ~0.045
from the numpy reference, the new one matches to ~1e-6. Small-sample PCA
is exactly the ill-conditioned regime the batched eigh bug (found while
working on QDA, #349) would be expected to hurt most.

Author: RicardoSantos-99

lib/scholar/decomposition/pca.ex

@@ -100,12 +100,12 @@ defmodule Scholar.Decomposition.PCA do
       iex> pca.components
       Nx.tensor(
         [
-          [0.36182016134262085, -0.08202514797449112, 0.8565111756324768, 0.3588128685951233],
-          [0.6585038900375366, 0.7275884747505188, -0.17632202804088593, -0.07679986208677292]
+          [0.36158934, -0.082268484, 0.8565728, 0.35884288],
+          [0.65654594, 0.72970587, -0.17576799, -0.074712664]
         ]
       )
       iex> pca.singular_values
-      Nx.tensor([25.089859008789062, 6.007821559906006])
+      Nx.tensor([25.089865, 6.007852])
   """
   deftransform fit(x, opts \\ []) do
     opts = NimbleOptions.validate!(opts, @opts_schema)
@@ -210,12 +210,12 @@ defmodule Scholar.Decomposition.PCA do
       iex> pca.components
       Nx.tensor(
         [
-          [-0.33354005217552185, 0.1048964187502861, -0.8618107080105579, -0.3674643635749817],
-          [-0.5862125754356384, -0.7916879057884216, 0.15874788165092468, -0.06621300429105759]
+          [-0.3331483, 0.105326004, -0.86190677, -0.36747244],
+          [-0.58288467, -0.79399943, 0.15797591, -0.06967129]
         ]
       )
       iex> pca.singular_values
-      Nx.tensor([77.05782028025969, 10.137848854064941])
+      Nx.tensor([77.05691, 10.158366])
   """
   deftransform incremental_fit(batches, opts) do
     opts = NimbleOptions.validate!(opts, @opts_schema)
@@ -242,12 +242,12 @@ defmodule Scholar.Decomposition.PCA do
     iex> pca.components
     Nx.tensor(
       [
-        [-0.3229745328426361, 0.09587063640356064, -0.8628664612770081, -0.37677285075187683],
-        [-0.6786625981330872, -0.7167785167694092, 0.14237160980701447, 0.07332050055265427]
+        [-0.32289746, 0.095940694, -0.8628819, -0.37678596],
+        [-0.67544055, -0.7201917, 0.141901, 0.070487075]
       ]
     )
     iex> pca.singular_values
-    Nx.tensor([166.141845703125, 6.078948020935059])
+    Nx.tensor([166.14182, 6.078579])
   """
   deftransform partial_fit(model, x) do
     if Nx.rank(x) != 2 do
@@ -373,9 +373,9 @@ defmodule Scholar.Decomposition.PCA do
       iex> Scholar.Decomposition.PCA.transform(pca, x_transform)
       Nx.tensor(
         [
-          [-1.4739344120025635, -0.48932668566703796],
-          [0.28113049268722534, 0.2337251454591751],
-          [2.0361955165863037, 0.9567767977714539]
+          [-1.4737695, -0.49078038],
+          [0.28105056, 0.23384926],
+          [2.0358706, 0.9584787]
         ]
       )
   """
@@ -442,13 +442,12 @@ defmodule Scholar.Decomposition.PCA do
       iex> Scholar.Decomposition.PCA.fit_transform(x, num_components: 2)
       Nx.tensor(
         [
-          [0.16441848874092102, 0.028548287227749825],
-          [-0.32804328203201294, 0.20709986984729767],
-          [-0.3284338414669037, -0.08318747580051422],
-          [-0.42237386107444763, -0.0735677033662796],
-          [0.17480169236660004, -0.11189625412225723],
-          [0.7396301627159119, 0.03300142288208008
-          ]
+          [0.16341472, -0.017230444],
+          [-0.33249766, 0.1893504],
+          [-0.32686582, -0.10110473],
+          [-0.42023695, -0.005525116],
+          [0.1768687, -0.14014886],
+          [0.7393165, 0.074657336]
         ]
       )
   """