Definition of Machine Learning :

Study of allowing computers to learn without explicit programming

A program is said to learn from a well-posed learning problem when:

Given some experience E, with respect to some task T and some performance measure P, A program is Learning if the performance of some task T, measured by some P improves with some experience E.

Types of Learning Algorithms

  • Supervised Learning
    • Correct answers are given explicitly
    • Model recieves feedback according to the answers that it predicted
    • Types of supervised learning tasks
      • Classification problem
        • Deals with discrete value outputs
        • May have multiple types of features to base predictions on (i.e. age, gender, race..)
      • Regression problem
        • Prediction of continuous value outputs
  • Unsupervised Learning
    • No labels or feedback given from the ‘teacher’
    • The point is to automatically find a categorization, structure, or patterns within the dataset
      • May open possibilities to further classification based on data
    • Cocktail party algorithm
        [W, s, v] = svd((repmat(sum(x.*x, 1), size(x, 1).1).*x)*x');
  • Misc
    • Reinforcement Learning
    • Recommender Learning