Replace utility classes with sklearn classes in NB4 #30
Conversation
PKGuo
changed the title
rosecers
left a comment
rosecers
left a comment
There was a problem hiding this comment.
I still argue that we should be using a precomputed kernel across the notebook. Recomputing the kernel with every model will greatly slow down the notebook.
| "from utilities.kernels import linear_kernel, gaussian_kernel, center_kernel\n", | ||
| "from utilities.classes import KPCA, KRR, SparseKPCA, SparseKRR\n", | ||
| "from utilities.classes import KPCovR, SparseKPCovR\n", | ||
| "from utilities.kernels import linear_kernel, gaussian_kernel, center_kernel # to be replaced with functions from skcosmo\n", |
There was a problem hiding this comment.
You can delete this if it is not used elsewhere.
There was a problem hiding this comment.
They are used somewhere.
There was a problem hiding this comment.
Please adapt this like you did in #29. There shouldn't be any kernel-tutorials kernel functions in this notebook. You should use sklearn kernels and centerers from skcosmo (scikit-learn-contrib/scikit-matter#8)
| "kernel_func = gaussian_kernel\n", | ||
| "kernel_type = 'gaussian'" |
There was a problem hiding this comment.
Are these used anywhere? If not, delete.
| "# Implementation in `skcosmo`\n", | ||
| "\n", | ||
| "Classes from the utility module enable computing SparseKPCA, SparseKRR, and SparseKPCovR with a scikit.learn-like syntax. `SparseKPCovR` is located in `utilities/kpcovr.py`." | ||
| "Classes from the skcosmo module enable computing SparseKPCA, SparseKRR, and SparseKPCovR with a scikit.learn-like syntax. `SparseKPCovR` is located in `skcosmo/PCovR/KPCovR.py`." |
Co-authored-by: rosecers <47536110+rosecers@users.noreply.github.com>
| "ref_kpca.fit(X_train, K=K_train)\n", | ||
| "\n", | ||
| "table_from_dict([ref_kpca.statistics(X_train, K=K_train), \n", | ||
| "ref_kpca = KernelPCA(n_components=n_PC, kernel='rbf',gamma = 1.0, fit_inverse_transform=True)\n", |
There was a problem hiding this comment.
See comment on line 364
rosecers
left a comment
rosecers
left a comment
There was a problem hiding this comment.
Two outstanding questions with respect to kernel definitions.
| "kernel_params = {\"kernel\": \"rbf\", \"gamma\": 1.0}\n", | ||
| "kernel_func = gaussian_kernel\n", | ||
| "kernel_params = {\"kernel\": \"precomputed\"}\n", | ||
| "kernel_func = rbf_kernel\n", |
There was a problem hiding this comment.
You can use functools to define
from functools import partial
kernel_func = partial(rbf_kernel, gamma=1.0)
to avoid having to repeat gamma everywhere below
rosecers
left a comment
rosecers
left a comment
There was a problem hiding this comment.
Please make sure this runs clean then feel free to merge
2 participants
Replace some utility classes in
4_SparseKernelMethods.ipynbwith appropriatesklearnclasses and modify the corresponding text.