Logistic Regression - Learning Notes & Implementations

A collection of notes and code implementations for understanding logistic regression and multiclass classification.

What's Inside

Theory Notes

• Logistic_regress.ipynb - Binary logistic regression fundamentals

  • Linear equation and sigmoid function
  • Log loss (binary cross-entropy)
  • Gradient descent optimization

• multiclass_logistic_regression.md - Multiclass classification approaches

  • One-vs-Rest (OvR)
  • Softmax function
  • Comparison between approaches

Code Implementations

Binary Logistic Regression

• logistic_regression.py - Simple binary classifier
  • Classifies points into two classes (0 or 1)
  • Uses gradient descent to optimize parameters
  • Demonstrates sigmoid function and log loss

Multiclass Classification

• multiclass.py - One-vs-Rest implementation

  • Example: Classifying animals (cat, dog, bird, fish)
  • Feature normalization
  • Binary classifier for one class (cats)

• softmax.py - Softmax implementation

  • Example: Classifying animals (cat, bird, fish)
  • One-hot encoding of labels
  • Categorical cross-entropy loss
  • Predicts all classes simultaneously