pytorch-linear-models

Implementation of linear models (OLS/LASSO/Ridge) in base PyTorch (so no torch.nn). This repo tries to follow the sklearn API for easy integration with existing projects.

Usage: Regression

from regression import LinearRegression

clf = LinearRegression(penalty=None)  # Penalty can be one of: None for  OLS, 'l1' for LASSO or 'l2' for Ridge

clf.fit(X_train,y_train)              # Fit the model like any sklearn model

clf.plot_history()                    # Plot loss over time

clf.predict(X_test)                   # Make predictions on new data

Usage: Classification

from classification import LogisticRegression

clf = LogisticRegression(penalty=None)# Penalty can be one of: None for  OLS, 'l1' for LASSO or 'l2' for Ridge

clf.fit(X_train,y_train)              # Fit the model like any sklearn model

clf.plot_history()                    # Plot loss over time

clf.predict(X_test)                   # Make predictions on new data

Requirements:

• torch==1.7.0
• numpy==1.18.5
• seaborn==0.10.0