https://github.com/khushi130404/regulexa

Regulexa is a Python project that showcases and compares Ridge, Lasso, and Elastic-Net regularization techniques in machine learning. It includes visualizations and performance insights to help prevent overfitting and improve model generalization.
https://github.com/khushi130404/regulexa

elastic-net-regression lasso-regression numpy ridge-regression

Last synced: 4 months ago
JSON representation

Regulexa is a Python project that showcases and compares Ridge, Lasso, and Elastic-Net regularization techniques in machine learning. It includes visualizations and performance insights to help prevent overfitting and improve model generalization.

• Host: GitHub
• URL: https://github.com/khushi130404/regulexa
• Owner: Khushi130404
• License: mit
• Created: 2024-12-26T08:56:50.000Z (6 months ago)
• Default Branch: main
• Last Pushed: 2024-12-29T03:16:06.000Z (6 months ago)
• Last Synced: 2025-01-12T04:55:01.202Z (5 months ago)
• Topics: elastic-net-regression, lasso-regression, numpy, ridge-regression
• Language: Jupyter Notebook
• Homepage:
• Size: 877 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Regulexa

Regulexa is a Python-based project designed to demonstrate and compare different regularization techniques used in machine learning: Ridge, Lasso, and Elastic-Net. Regularization helps to prevent overfitting and improves the generalization of models by adding a penalty term to the loss function. This project includes visualizations and performance comparisons for these techniques, making it a valuable resource for data science enthusiasts and machine learning practitioners.

## Regularization Techniques

1. Ridge Regression

- Ridge regression adds a penalty equal to the square of the magnitude of coefficients. It helps to reduce model complexity and multicollinearity.

- Penalty: α * ||w||₂² (L2 norm)

- Shrinks coefficients towards zero but never sets them exactly to zero.

2. Lasso Regression

- Lasso regression adds a penalty equal to the absolute value of the coefficients. It performs both variable selection and regularization.

- Penalty: α * ||w||₁ (L1 norm)

- Can shrink some coefficients to exactly zero, effectively performing feature selection.

3. Elastic-Net Regression

- Elastic-Net combines both Ridge and Lasso penalties.

- Penalty: α * [(1 - λ) ||w||₂² + λ ||w||₁]

- Suitable for datasets with correlated features and when feature selection is required.

## Features

- Synthetic Data: Generates synthetic datasets for demonstration purposes.

- Visualizations: Plots showing how regularization affects coefficients and model performance.

- Comparisons: Side-by-side comparison of Ridge, Lasso, and Elastic-Net.

- Metrics: Evaluation using metrics like Mean Squared Error (MSE) and R².

## Tech Used

- Jupyter Notebook: For interactive coding and visualizations.

- scikit-learn (sklearn): For implementing regularization techniques and machine learning models.

- Matplotlib: For creating visualizations.

- NumPy: For numerical computations.

- Pandas: For data manipulation and analysis.