Chapter 1

1. Applications and problems

Applications

• Text or document classification, e.g., spam detection;
• Natural language processing, e.g., morphological analysis, part-of-speech tagging, statistical parsing, named-entity recognition;
• Speechrecognition, speech synthesis, speaker verification;
• Optical character recognition (OCR);
• Computational biology applications, e.g., protein function or structured prediction;
• Computer vision tasks, e.g., image recognition, face detection;
• Faud detection (credit card, telephone) and network intrusion;
• Games, e.g., chess, backgammon;
• Medical diagnosis;
• Recommendation systems, search enginesm information extraction systems.

Problems

• Classification
• Regression
• Ranking
• Clustering
• Dimensionality reduction or manifold learning

1.2 Definitions and terminology

• Examples
• Features
• Labels
• Training sample
• Validation sample
• Test sample
• Loss function
• Hypothesis set

1.3 Cross-validation

In practice, the amount of labeled data available is often too small to set aside a validation sample since that would leave an insufficient amount of training data. Instead, a widely adopted method known as n-fold cross-validation is used to exploit the labeled data both for model selection (selection of the free parameters of the algorithm) and for training.

 1.4 Learning scenarios

• Supervised learning
  • The learner receives a set of labeled examples as training data and makes predictions for all unseen points.
• Unsupervised learning
  • The learner exclusively receives unlabeled training data,and makes predictions for all unseen points.
• Semi-unsupervised learning
  • The learner receives a training sample consisting of both labeled and unlabeled data, and makes predictions for all unseen points.
• Transductive inference
  • As in the semi-supervised scenario, the learner receives a labeled training sample along with a set of unlabeled test points. However, the objective of transductive inference is to predict labels only for these particular test points.
• On-line learning
  • In contrast with the previous scenarios, the online scenario

    involves multiple rounds and training and testing phases are intermixed. At each

    round, the learner receives an unlabeled training point, makes a prediction, receives 

    the true label, and incurs a loss

• Reinforcement learning
  • The training and testing phases are also intermixed in reinforcement learning. To collect information, the learner actively interacts with the environment and in some cases affects the environment, and receives an immediate reward for each action. The object of the learner is to maximize his reward over a course of actions and iterations with the environment.
• Active learning

  • 　　The learner adaptively or interactively collects training examples,
    typically by querying an oracle to request labels for new points.