Kernel PCA: Python and Benchmarking Code #5988
Open
+902
−5
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
tomasjoh
changed the title
tomasjoh
commented
Jul 27, 2024
| def _build_params(self, n_rows, n_cols): | ||
| IF GPUBUILD == 1: | ||
| cdef paramsKPCA *params = new paramsKPCA() | ||
| params.n_components = min(self.n_components_ or n_rows, n_rows) |
There was a problem hiding this comment.
Sklearn reference for setting n_components
11 tasks
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
This PR relies on C++ implementation from #5987
Adds Python, and benchmarking code for Kernel PCA. This implementation of Kernel PCA support fit(), transform(), and fit_transform().
Feature request: #1317
Tests and benchmarks were performed on an EC2
g4dn.xlargeinstance with CUDA 12.2.Click here to see environment details
Notes for Reviewers
The API deviates from SKlearn by not supporting options for these fields: fit_inverse_transform, random_state, n_jobs, max_iter. If a user tries to set one of them a
NotImplementedErrorwill be raised.The Criteria of Done mentions making the class pickable in
cuml/tests/test_pickle.py. I couldn't find a PCA reference for this. Would appreciate pointers if additional work is needed.Benchmarks
From
notebooks/tools/cuml_benchmarks.ipynbBenchmark output
We see an even greater speedup when we set n_components = n_samples. Setting n_components to n_samples is the same as default behavior, except zero eigenvalues aren't removed.
Manual tests
Kernel PCA with RBF kernel
code
Kernel PCA with poly kernel
code
from sklearn.datasets import make_classification def plot_3d_projection(X, y, title, elev=30, azim=30): fig = plt.figure(figsize=(8, 6)) ax = fig.add_subplot(111, projection='3d') for target in np.unique(y): ax.scatter(X[y == target, 0], X[y == target, 1], X[y == target, 2], label=str(target)) ax.set_title(title) ax.set_xlabel('Component 1') ax.set_ylabel('Component 2') ax.set_zlabel('Component 3') ax.legend() ax.view_init(elev=elev, azim=azim) # Set viewpoint plt.show() X, y = make_classification(n_features=3, n_informative=3, n_redundant=0, n_clusters_per_class=1, n_classes=3) plot_3d_projection(X, y, 'Original Data', elev=30, azim=60) poly_kpca = cuKernelPCA(n_components=3, kernel='poly', degree=5, gamma=2, coef0=2) X_poly_kpca = linear_kpca.fit_transform(X_linear) plot_3d_projection(X_poly_kpca, y, 'cuML KernelPCA with Poly Kernel', elev=30, azim=120)Projecting testing data
Case is copied from sklearn, except it uses cuML PCA and kernelPCA
code
Definition of Done Criteria Checklist
Python Checklist
Design
cuml/tests/test_pickle.pyinput_to_cuml_arrayto accept flexible inputs and check their datatypes and usecumlArray.to_output()to return configurable outputs.CumlArrayTesting
python/cuml/benchmarks/algorithms.pyand benchmarks notebook inpython/cuml/notebooks/tools/cuml_benchmarks.ipynbUnit test results
Python Test Results