Machine learning
Machine learning (ML) is a field of artificial intelligence concerned with developing computational methods that enable systems to learn patterns, relationships, and representations from data and use them to make predictions, classifications, decisions, or generate outputs without relying exclusively on explicitly programmed rules.
Machine learning systems are trained using data and mathematical algorithms to identify patterns that can be applied to new or previously unseen inputs. The field is widely used in applications including natural language processing, computer vision, speech recognition, recommendation systems, fraud detection, forecasting, robotics, scientific research, and generative AI.
Machine learning includes several major approaches, including supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. Modern machine learning increasingly relies on Deep learning, which uses multilayer artificial neural networks to learn complex representations from large datasets.
Objectives
The principal objectives of machine learning are to:
- learn patterns from data;
- make predictions;
- classify or organize information;
- identify relationships within datasets;
- automate decision-making;
- generate useful outputs from new data;
- improve system performance through experience or additional training.
Major types
Machine learning is commonly divided into several major approaches:
- supervised learning;
- unsupervised learning;
- semi-supervised learning;
- reinforcement learning.
Supervised learning
Supervised learning involves training a model using labeled data in which the desired output is known.
The model learns a relationship between input variables and corresponding outputs and uses that relationship to make predictions on new data.
Common applications include:
- classification;
- regression;
- forecasting;
- image recognition;
- fraud detection.
Unsupervised learning
Unsupervised learning involves training models using data without predefined output labels.
The objective is generally to identify patterns, structures, clusters, or relationships within the data.
Common applications include:
- clustering;
- dimensionality reduction;
- anomaly detection;
- customer segmentation.
Semi-supervised learning
Semi-supervised learning combines labeled and unlabeled data during model training.
It can be useful when obtaining large quantities of labeled training data is expensive or difficult.
Reinforcement learning
Reinforcement learning is an approach in which an agent learns to make decisions by interacting with an environment and receiving rewards or penalties based on its actions.
The objective is generally to learn a policy that maximizes cumulative reward.
Applications include robotics, game playing, optimization, and autonomous systems.
Algorithms
Machine learning uses a wide range of algorithms and statistical techniques, including:
- linear regression;
- logistic regression;
- decision trees;
- random forests;
- support vector machines;
- k-means clustering;
- gradient boosting;
- neural networks.
The appropriate algorithm depends on the characteristics of the data and the intended application.
Training
Training is the process through which a machine learning model learns from a dataset.
During training, model parameters are adjusted to reduce the difference between predicted and expected outcomes or to optimize another specified objective.
A dataset is commonly divided into:
- training data;
- validation data;
- test data.
The training dataset is used to fit the model, while validation and test datasets are used to evaluate its performance.
Overfitting and underfitting
Overfitting occurs when a model learns the training data too closely and performs poorly on new data.
Underfitting occurs when a model is insufficiently complex to capture important patterns in the data.
Machine learning development seeks an appropriate balance between model complexity and generalization.
Model evaluation
Machine learning models are evaluated using metrics appropriate to the task.
Common metrics include:
- accuracy;
- precision;
- recall;
- F1 score;
- mean squared error;
- mean absolute error;
- area under the receiver operating characteristic curve.
The appropriate evaluation method depends on the problem, dataset, and consequences of prediction errors.
Deep learning
Deep learning is a major branch of machine learning based on multilayer artificial neural networks.
Deep-learning models can learn complex representations from large datasets and have produced significant advances in areas such as computer vision, speech recognition, natural language processing, and generative AI.
Neural networks
Artificial neural networks are computational models inspired in part by the structure of biological neural systems.
They consist of interconnected computational units organized into layers. During training, the parameters of these networks are adjusted to improve performance on a specified task.
Applications
Machine learning is used in:
- Natural language processing;
- computer vision;
- speech recognition;
- recommendation systems;
- financial analysis;
- fraud detection;
- cybersecurity;
- medical research;
- scientific computing;
- autonomous vehicles;
- robotics;
- forecasting;
- search engines;
- business intelligence;
- Generative artificial intelligence.
Relationship with artificial intelligence
Artificial intelligence is the broader field concerned with developing systems capable of performing tasks associated with human intelligence.
Machine learning is a major approach within artificial intelligence and provides many of the methods used to develop modern AI systems.
Relationship with natural language processing
Natural language processing applies computational methods to human language.
Machine learning provides many of the algorithms used in modern NLP systems, including systems for classification, translation, speech processing, information extraction, and language generation.
Relationship with large language models
Large language models are generally developed using machine learning and deep learning techniques.
Their training involves optimizing large numbers of model parameters using extensive datasets and computational resources.
Relationship with generative artificial intelligence
Generative artificial intelligence uses machine learning models to generate new content such as text, images, audio, video, and software code.
Machine learning therefore provides many of the underlying methods used by modern generative AI systems.
Limitations
Machine learning systems may produce inaccurate predictions or decisions because of:
- insufficient training data;
- poor-quality data;
- biased datasets;
- inappropriate model selection;
- distribution changes;
- overfitting;
- measurement errors.
Other challenges include computational requirements, interpretability, privacy, security, and the difficulty of establishing why some complex models produce particular outputs.
Ethical considerations
The use of machine learning raises issues concerning:
- algorithmic bias;
- privacy;
- data protection;
- transparency;
- accountability;
- discrimination;
- security;
- automated decision-making.
The appropriate governance of machine learning systems depends on the application, potential risks, and applicable legal and regulatory requirements.
Importance
Machine learning is a foundational technology of modern artificial intelligence because it enables computational systems to learn from data rather than relying solely on explicitly programmed rules. Its application across language processing, computer vision, scientific research, finance, cybersecurity, robotics, and generative AI has made it one of the principal methods used to develop modern intelligent systems.