Kernel PCA: Python and Benchmarking Code

BUILD.md

@@ -91,7 +91,7 @@ $ pytest --ignore=cuml/tests/dask --ignore=cuml/tests/test_nccl.py
 
 If you want a list of the available Python tests:
 ```bash
-$ pytest cuML/tests --collect-only
+$ pytest cuml/tests --collect-only
 ```
 
 ### Manual Process

notebooks/tools/cuml_benchmarks.ipynb

@@ -531,6 +531,30 @@
     "execute_benchmark(\"PCA\", runner)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### KernelPCA<a id=\"kernel_pca\"/>"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "runner = cuml.benchmark.runners.SpeedupComparisonRunner(\n",
+    "    bench_rows=[400, 800, 1600, 3200, 6400, 12800],\n",
+    "    bench_dims=WIDE_FEATURES,\n",
+    "    dataset_name=DATA_CLASSIFICATION,\n",
+    "    input_type=INPUT_TYPE,\n",
+    "    n_reps=N_REPS\n",
+    ")\n",
+    "\n",
+    "execute_benchmark(\"KernelPCA\", runner)"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},

python/cuml/CMakeLists.txt

@@ -109,6 +109,7 @@ if(CUML_CPU)
 
   set(CUML_ALGORITHMS "linearregression")
   list(APPEND CUML_ALGORITHMS "pca")
+  list(APPEND CUML_ALGORITHMS "kpca")
   list(APPEND CUML_ALGORITHMS "tsvd")
   list(APPEND CUML_ALGORITHMS "elasticnet")
   list(APPEND CUML_ALGORITHMS "logisticregression")
@@ -179,6 +180,7 @@ add_subdirectory(cuml/svm)
 add_subdirectory(cuml/tsa)
 
 add_subdirectory(cuml/experimental/linear_model)
+add_subdirectory(cuml/experimental/decomposition)
 
 if(DEFINED cython_lib_dir)
   rapids_cython_add_rpath_entries(TARGET cuml PATHS "${cython_lib_dir}")

python/cuml/cuml/benchmark/algorithms.py

@@ -245,6 +245,13 @@ def all_algorithms():
             name="PCA",
             accepts_labels=False,
         ),
+        AlgorithmPair(
+            sklearn.decomposition.KernelPCA,
+            cuml.experimental.KernelPCA,
+            shared_args=dict(),
+            name="KernelPCA",
+            accepts_labels=False,
+        ),
         AlgorithmPair(
             sklearn.decomposition.TruncatedSVD,
             cuml.decomposition.tsvd.TruncatedSVD,

python/cuml/cuml/decomposition/utils.pxd

@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2019-2022, NVIDIA CORPORATION.
+# Copyright (c) 2019-2024, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -18,6 +18,19 @@ from libcpp cimport bool
 
 ctypedef int underlying_type_t_solver
 
+cdef extern from "raft/distance/distance_types.hpp" namespace "raft::distance::kernels" nogil:
+    enum KernelType:
+        LINEAR,
+        POLYNOMIAL,
+        RBF,
+        TANH
+
+    cdef struct KernelParams:
+        KernelType kernel
+        int degree
+        double gamma
+        double coef0
+
 cdef extern from "cuml/decomposition/params.hpp" namespace "ML" nogil:
 
     ctypedef enum solver "ML::solver":
@@ -41,3 +54,11 @@ cdef extern from "cuml/decomposition/params.hpp" namespace "ML" nogil:
     cdef cppclass paramsPCA(paramsTSVD):
         bool copy
         bool whiten
+
+    cdef cppclass paramsKPCA(paramsTSVD):
+        KernelParams kernel
+        size_t n_training_samples
+        bool copy
+        bool remove_zero_eig
+        bool fit_inverse_transform
+

python/cuml/cuml/experimental/__init__.py

@@ -1 +1,2 @@
 from cuml.experimental.fil import ForestInference
+from cuml.experimental.decomposition import KernelPCA

python/cuml/cuml/experimental/decomposition/CMakeLists.txt

@@ -0,0 +1,26 @@
+# =============================================================================
+# Copyright (c) 2024, NVIDIA CORPORATION.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except
+# in compliance with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software distributed under the License
+# is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express
+# or implied. See the License for the specific language governing permissions and limitations under
+# the License.
+# =============================================================================
+
+
+
+set(cython_sources "")
+add_module_gpu_default("kpca.pyx" ${kpca_algo} ${decomposition_algo})
+
+rapids_cython_create_modules(
+  CXX
+  SOURCE_FILES "${cython_sources}"
+  LINKED_LIBRARIES "${cuml_sg_libraries}"
+  MODULE_PREFIX experimental_
+  ASSOCIATED_TARGETS cuml
+)

python/cuml/cuml/experimental/decomposition/__init__.py

@@ -0,0 +1,17 @@
+#
+# Copyright (c) 2024, NVIDIA CORPORATION.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+from cuml.experimental.decomposition.kpca import KernelPCA