Machine Learning is a branch of Artificial Intelligence concerned with algorithms that learn patterns from data and improve their performance through experience.
Learning from labelled input–output data.
Sub-branches:
Classification
Binary classification
Multiclass classification
Multilabel classification
Ordinal classification
Regression
Linear regression
Polynomial regression
Logistic regression
Support vector machines
Decision trees
Random forests
Gradient-boosting methods
Nearest-neighbour methods
Bayesian classifiers
Learning-to-rank
Learning patterns from unlabelled data.
Sub-branches:
Clustering
K-means clustering
Hierarchical clustering
Density-based clustering
Gaussian mixture models
Dimensionality reduction
Principal component analysis
Independent component analysis
Matrix factorization
Manifold learning
Association-rule learning
Topic modelling
Anomaly detection
Density estimation
Latent-variable modelling
Learning from a small amount of labelled data and a large amount of unlabelled data.
Sub-branches:
Self-training
Co-training
Pseudo-labelling
Graph-based semi-supervised learning
Consistency regularization
Entropy minimization
Semi-supervised generative modelling
Learning representations by generating training signals directly from unlabelled data.
Sub-branches:
Contrastive learning
Masked-data modelling
Predictive representation learning
Autoencoding
Pretext-task learning
Siamese-network learning
Multimodal self-supervised learning
Language-model pretraining
Image representation learning
Audio representation learning
Learning through interaction with an environment using rewards and penalties.
Sub-branches:
Model-free reinforcement learning
Model-based reinforcement learning
Value-based learning
Policy-based learning
Actor–critic methods
Q-learning
Deep Q-networks
Temporal-difference learning
Monte Carlo reinforcement learning
Multi-agent reinforcement learning
Hierarchical reinforcement learning
Offline reinforcement learning
Inverse reinforcement learning
Safe reinforcement learning
Meta-reinforcement learning
Robotics reinforcement learning
Machine learning using multilayer artificial neural networks.
Sub-branches:
Feedforward neural networks
Convolutional neural networks
Recurrent neural networks
Long short-term memory networks
Gated recurrent units
Transformer networks
Graph neural networks
Autoencoders
Variational autoencoders
Generative adversarial networks
Diffusion models
Deep belief networks
Neural ordinary differential equations
Capsule networks
Spiking neural networks
Deep reinforcement learning
Automatically learning useful features or representations from raw data.
Sub-branches:
Feature learning
Embedding learning
Metric learning
Manifold representation learning
Disentangled representation learning
Sparse representation learning
Distributed representations
Multimodal representation learning
Contrastive representation learning
Latent-space modelling
Learning the statistical distribution of data to generate new content.
Sub-branches:
Generative adversarial networks
Variational autoencoders
Autoregressive models
Diffusion models
Normalizing flows
Energy-based models
Generative language models
Image-generation models
Audio-generation models
Video-generation models
Synthetic-data generation
Text-to-image modelling
Multimodal generation
Using probability theory to model uncertainty and data-generating processes.
Sub-branches:
Bayesian machine learning
Bayesian networks
Gaussian processes
Hidden Markov models
Markov random fields
Probabilistic graphical models
Bayesian deep learning
Monte Carlo methods
Variational inference
Expectation-maximization
Probabilistic programming
Uncertainty quantification
Applying statistical theory to prediction, inference and pattern discovery.
Sub-branches:
Parametric learning
Non-parametric learning
Statistical classification
Statistical regression
Regularization methods
Kernel methods
Bayesian inference
Hypothesis-based learning
Empirical risk minimization
Statistical decision theory
Bias–variance analysis
Generalization theory
Combining several models to improve predictive performance.
Sub-branches:
Bagging
Boosting
Stacking
Voting
Random forests
AdaBoost
Gradient boosting
XGBoost
LightGBM
CatBoost
Bayesian model averaging
Mixture-of-experts models
Updating a model continuously as new data arrive.
Sub-branches:
Online learning
Incremental learning
Streaming-data learning
Continual learning
Lifelong learning
Adaptive learning
Concept-drift detection
Real-time learning
Online optimization
Memory-based continual learning
Selecting the most informative data samples for human labelling.
Sub-branches:
Pool-based active learning
Stream-based active learning
Query synthesis
Uncertainty sampling
Query-by-committee
Expected-error reduction
Diversity-based sampling
Human-in-the-loop learning
Using knowledge learned in one task or domain to improve another.
Sub-branches:
Feature-transfer learning
Fine-tuning
Domain adaptation
Domain generalization
Cross-domain learning
Cross-lingual transfer learning
Multitask transfer
Zero-shot learning
One-shot learning
Few-shot learning
Foundation-model adaptation
Learning several related tasks simultaneously.
Sub-branches:
Hard parameter sharing
Soft parameter sharing
Joint representation learning
Auxiliary-task learning
Multi-objective learning
Cross-task transfer
Multimodal multitask learning
Learning how to learn efficiently from new tasks or limited data.
Sub-branches:
Optimization-based meta-learning
Metric-based meta-learning
Model-based meta-learning
Few-shot meta-learning
Meta-reinforcement learning
Neural architecture meta-learning
Hyperparameter meta-learning
Training models across multiple devices or locations without centralizing all data.
Sub-branches:
Federated learning
Horizontal federated learning
Vertical federated learning
Cross-device federated learning
Cross-silo federated learning
Distributed deep learning
Decentralized learning
Split learning
Communication-efficient learning
Privacy-preserving federated learning
Protecting sensitive information during model training and inference.
Sub-branches:
Differential privacy
Secure multiparty computation
Homomorphic encryption
Federated privacy
Private inference
Data anonymization
Synthetic-data privacy
Confidential machine learning
Privacy-preserving data mining
Understanding how and why models produce particular results.
Sub-branches:
Intrinsically interpretable models
Post-hoc explanation
Local explanation
Global explanation
Feature-attribution methods
Counterfactual explanations
Concept-based explanations
Model visualization
Explainable deep learning
Causal explanations
Learning cause-and-effect relationships rather than only correlations.
Sub-branches:
Causal inference
Causal discovery
Structural causal models
Treatment-effect estimation
Counterfactual reasoning
Causal representation learning
Instrumental-variable learning
Causal reinforcement learning
Causal transfer learning
Learning from graph-structured data.
Sub-branches:
Graph neural networks
Graph convolutional networks
Graph attention networks
Graph representation learning
Node classification
Link prediction
Graph classification
Knowledge-graph learning
Graph generation
Geometric deep learning
Dynamic graph learning
Heterogeneous graph learning
Learning from non-Euclidean structures such as graphs, manifolds and meshes.
Sub-branches:
Graph-based learning
Manifold learning
Mesh learning
Point-cloud learning
Spherical neural networks
Equivariant neural networks
Topological deep learning
Lie-group learning
Learning from data recorded over time.
Sub-branches:
Time-series forecasting
Sequence classification
Temporal anomaly detection
State-space modelling
Temporal clustering
Change-point detection
Survival analysis
Event prediction
Spatiotemporal modelling
Neural time-series modelling
Machine-learning methods for human language.
Sub-branches:
Natural language processing
Language modelling
Text classification
Sentiment analysis
Machine translation
Question answering
Information extraction
Named-entity recognition
Text summarization
Dialogue systems
Speech-language modelling
Retrieval-augmented generation
Large language models
Machine-learning methods for images and video.
Sub-branches:
Image classification
Object detection
Image segmentation
Facial recognition
Pose estimation
Optical character recognition
Medical-image analysis
Video understanding
Visual tracking
Image restoration
Image generation
Three-dimensional vision
Remote-sensing image analysis
Learning from speech, music and other audio signals.
Sub-branches:
Automatic speech recognition
Speaker recognition
Speech synthesis
Voice conversion
Audio classification
Music information retrieval
Sound-event detection
Acoustic-scene analysis
Speech enhancement
Audio generation
Learning jointly from multiple data types.
Sub-branches:
Vision-language learning
Audio-visual learning
Text-audio learning
Sensor-fusion learning
Cross-modal retrieval
Multimodal generation
Multimodal representation learning
Multimodal reasoning
Vision-language-action models
Predicting items, services or content relevant to users.
Sub-branches:
Collaborative filtering
Content-based recommendation
Hybrid recommendation
Context-aware recommendation
Sequential recommendation
Session-based recommendation
Knowledge-based recommendation
Graph-based recommendation
Social recommendation
Reinforcement-learning recommendation
Identifying unusual, rare or previously unseen patterns.
Sub-branches:
Statistical anomaly detection
Density-based anomaly detection
Distance-based anomaly detection
Isolation-based methods
One-class classification
Deep anomaly detection
Time-series anomaly detection
Fraud detection
Outlier detection
Novelty detection
Using evolutionary computation to design or optimize learning systems.
Sub-branches:
Genetic algorithms
Genetic programming
Evolution strategies
Neuroevolution
Swarm intelligence
Particle-swarm optimization
Ant-colony optimization
Evolutionary neural architecture search
Multi-objective evolutionary learning
Automating the construction and optimization of machine-learning systems.
Sub-branches:
Automated data preprocessing
Automated feature engineering
Algorithm selection
Hyperparameter optimization
Neural architecture search
Model selection
Pipeline optimization
Automated ensemble construction
Automated deployment
Automated model monitoring
Developing models that remain reliable under noise, attacks or distribution changes.
Sub-branches:
Adversarial attacks
Adversarial training
Robust optimization
Distributionally robust learning
Out-of-distribution detection
Certified robustness
Data-poisoning defence
Model-stealing defence
Backdoor-attack detection
Reliable machine learning
Developing machine-learning systems that are ethical, accountable and socially responsible.
Sub-branches:
Algorithmic fairness
Bias detection
Bias mitigation
Model accountability
AI transparency
Responsible data governance
Ethical machine learning
Human-centred machine learning
AI auditing
Societal-impact assessment
Running machine-learning models on resource-constrained devices.
Sub-branches:
Edge AI
TinyML
On-device learning
Model compression
Quantization
Pruning
Knowledge distillation
Efficient neural networks
Hardware-aware learning
Embedded machine learning
Combining quantum computing and machine-learning methods.
Sub-branches:
Quantum supervised learning
Quantum unsupervised learning
Quantum reinforcement learning
Quantum neural networks
Variational quantum algorithms
Quantum kernel methods
Quantum generative models
Quantum-enhanced optimization
Classical–quantum hybrid learning
Applying machine learning to scientific modelling, simulation and discovery.
Sub-branches:
Physics-informed neural networks
Neural differential equations
Surrogate modelling
Equation discovery
Symbolic regression
Simulation-based inference
Data-driven physical modelling
Climate machine learning
Computational biology learning
Materials informatics
Astronomy and astrophysics machine learning
Particle-physics machine learning
Engineering practices for deploying and maintaining machine-learning systems.
Sub-branches:
Data pipelines
Model deployment
Model serving
Model versioning
Experiment tracking
Data versioning
Model monitoring
Drift detection
Continuous training
Continuous integration and deployment
Feature stores
ML system governance
Mathematical study of learning algorithms and their limitations.
Sub-branches:
Computational learning theory
Statistical learning theory
PAC learning
VC-dimension theory
Generalization theory
Online learning theory
Optimization theory
Information-theoretic learning
Algorithmic stability
Sample-complexity analysis
Learning bounds
Kernel theory
The broadest machine-learning classification can be summarized as:
Supervised learning
Unsupervised learning
Semi-supervised learning
Self-supervised learning
Reinforcement learning
Deep learning
Generative learning
Transfer and meta-learning
Probabilistic and statistical learning
Graph and geometric learning
Multimodal and domain-specific learning
Responsible, secure and privacy-preserving learning
Scientific and theoretical machine learning
Automated ML and MLOps