A Functional Taxonomy of Common Machine Learning Algorithms
Summary
This overview organizes machine learning algorithms in two ways: by learning setup and by functional family. It distinguishes supervised learning with labeled examples, unsupervised learning that finds structure in unlabeled data, and semi-supervised learning that uses both. Its main catalog groups methods such as regression, instance-based learning, regularization, decision trees, Bayesian methods, clustering, association rules, neural networks, dimensionality reduction, and ensembles, with familiar examples in each group.
The document explains broad purposes rather than giving implementation steps, comparisons, or trading results. For example, it describes instance-based prediction as matching new observations against stored examples, and ensembles as combining predictions from separately trained models. The classifications are explicitly incomplete and sometimes subjective: algorithms can fit multiple families, and terms such as regression can refer to either a task or a method. Specialized fields are also omitted. The trading references appear only as titles of related quantitative stock-selection research, without details or evidence about those strategies.
Key ideas
- Machine learning can be grouped by whether training data are labeled, unlabeled, or a mixture of both.
- Functional families include regression, decision trees, Bayesian methods, clustering, neural networks, and ensembles.
- Instance-based methods compare new observations with stored examples to make predictions.
- Regularization penalizes model complexity to encourage simpler models that may generalize better.
- The taxonomy is illustrative rather than exhaustive, and some methods plausibly belong to multiple groups.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.