User Guide

• 1. What is Metric Learning?
  • 1.1. Problem Setting
  • 1.2. Mahalanobis Distances
  • 1.3. Use-cases
  • 1.4. Further reading
• 2. Supervised Metric Learning
  • 2.1. General API
    • 2.1.1. Input data
    • 2.1.2. Fit, transform, and so on
    • 2.1.3. Scikit-learn compatibility
  • 2.2. Algorithms
    • 2.2.1. LMNN
    • 2.2.2. NCA
    • 2.2.3. LFDA
    • 2.2.4. MLKR
    • 2.2.5. Supervised versions of weakly-supervised algorithms
• 3. Weakly Supervised Metric Learning
  • 3.1. General API
    • 3.1.1. Input data
      • 3.1.1.1. Basic form
      • 3.1.1.2. 3D array of tuples
      • 3.1.1.3. 2D array of indicators + preprocessor
    • 3.1.2. Fit, transform, and so on
    • 3.1.3. Prediction and scoring
    • 3.1.4. Scikit-learn compatibility
  • 3.2. Learning on pairs
    • 3.2.1. Fitting
    • 3.2.2. Prediction
      • 3.2.2.1. Prediction threshold
    • 3.2.3. Scoring
    • 3.2.4. Algorithms
      • 3.2.4.1. ITML
      • 3.2.4.2. SDML
      • 3.2.4.3. RCA
      • 3.2.4.4. MMC
  • 3.3. Learning on triplets
    • 3.3.1. Fitting
    • 3.3.2. Prediction
    • 3.3.3. Scoring
    • 3.3.4. Algorithms
      • 3.3.4.1. SCML
  • 3.4. Learning on quadruplets
    • 3.4.1. Fitting
    • 3.4.2. Prediction
    • 3.4.3. Scoring
    • 3.4.4. Algorithms
      • 3.4.4.1. LSML
• 4. Unsupervised Metric Learning
  • 4.1. Algorithms
    • 4.1.1. Covariance
• 5. Preprocessor
  • 5.1. Array-like
  • 5.2. Callable