Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/jameschapman19/scikit-prox
A package for fitting regularized models from scikit-learn via proximal gradient descent
https://github.com/jameschapman19/scikit-prox
proximal-gradient-descent regularization scikit-learn scikit-learn-api
Last synced: 4 days ago
JSON representation
A package for fitting regularized models from scikit-learn via proximal gradient descent
- Host: GitHub
- URL: https://github.com/jameschapman19/scikit-prox
- Owner: jameschapman19
- License: mit
- Created: 2023-01-27T15:31:37.000Z (over 1 year ago)
- Default Branch: master
- Last Pushed: 2023-06-28T13:50:44.000Z (over 1 year ago)
- Last Synced: 2024-09-26T00:10:43.264Z (8 days ago)
- Topics: proximal-gradient-descent, regularization, scikit-learn, scikit-learn-api
- Language: Python
- Homepage: https://scikit-prox.readthedocs.io/en/latest/
- Size: 104 KB
- Stars: 4
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Contributing: docs/contributing.rst
- License: LICENSE.md
Awesome Lists containing this project
README
[](https://codecov.io/gh/jameschapman19/scikit-prox)
[](https://scikit-prox.readthedocs.io/en/latest/?badge=latest)
[](https://pypi.org/project/scikit-prox/)
[](https://pypi.org/project/scikit-prox/)
# Scikit-Prox
The goal of this project is to implement a set of algorithms for solving the following optimization problem:
minimize f(x) + g(x) where f is a smooth function and g is a proximal operator. The proximal operator of a function g is defined as:
proxg(λx) = argmin y g(y) + 1/2λ‖y − x‖2
## Installation
To install the package, run the following command:
pip install scikit-prox
## Usage
### Example 1: Lasso
The following code solves the following optimization problem:
minimize 1/2‖Ax − b‖2 + λ‖x‖1
```python
import numpy as np
from scipy import sparse
from sklearn.datasets import make_regression
from sklearn.linear_model import Lasso
from skprox.linear_model import RegularisedLinearRegression
# Generate data
X, y = make_regression(n_samples=100, n_features=1000, random_state=0, noise=4.0, bias=100.0)
X = sparse.csr_matrix(X)
# Solve the problem using scikit-learn
model = Lasso(alpha=0.1)
model.fit(X, y)
print("scikit-learn solution: {}".format(model.coef_))
# Solve the problem using scikit-prox
model = RegularisedLinearRegression(proximal='L1', sigma=0.1)
model.fit(X, y)
print("scikit-prox solution: {}".format(model.coef_))
```
### Example 2: Total Variation Regression
The following code solves the following optimization problem:
minimize 1/2‖Ax − b‖2 + λ‖∇x‖1
```python
import numpy as np
from scipy import sparse
from sklearn.datasets import make_regression
from skprox.linear_model import RegularisedLinearRegression
# Generate data
X, y = make_regression(n_samples=100, n_features=1000, random_state=0, noise=4.0, bias=100.0)
X = sparse.csr_matrix(X)
# Solve the problem using scikit-prox
model = RegularisedLinearRegression(proximal='TV', sigma=0.1)
model.fit(X, y)
print("scikit-prox solution: {}".format(model.coef_))
```
### Example 3: Grid Search
The following code solves the following optimization problem:
minimize 1/2‖Ax − b‖2 + λ‖x‖1
```python
import numpy as np
from scipy import sparse
from sklearn.datasets import make_regression
from sklearn.linear_model import Lasso
from skprox.linear_model import RegularisedLinearRegression
from sklearn.model_selection import GridSearchCV
# Generate data
X, y = make_regression(n_samples=100, n_features=1000, random_state=0, noise=4.0, bias=100.0)
X = sparse.csr_matrix(X)
# Solve the problem using scikit-learn
model = Lasso()
grid = GridSearchCV(model, {'alpha': [0.1, 0.2, 0.3]})
grid.fit(X, y)
print("scikit-learn solution: {}".format(grid.best_estimator_.coef_))
# Solve the problem using scikit-prox
model = RegularisedLinearRegression(proximal='L1')
grid = GridSearchCV(model, {'sigma': [0.1, 0.2, 0.3]})
grid.fit(X, y)
print("scikit-prox solution: {}".format(grid.best_estimator_.coef_))
```
## Documentation
The documentation is available at https://scikit-prox.readthedocs.io/en/latest/
## License
This project is licensed under the MIT License - see the LICENSE.md file for details
## Acknowledgments
This project leans on the pyproximal package borrowing all the proximal operators except for Total Variation which
is implemented using functions from skimage.