Machine Learning (ML) is a subset of artificial intelligence (AI) focused on building systems that learn from data to make and improve decisions or predictions. Unlike traditional software, which follows explicit instructions written by programmers, ML systems are trained using large sets of data and algorithms that adjust themselves to perform better as they process more data.
Key Concepts in Machine Learning:
- Supervised Learning: The algorithm learns from labeled training data, trying to predict outcomes for new, unseen data based on this training.
- Unsupervised Learning: The algorithm explores input data without labeled responses with the goal of finding structure in the data (e.g., clustering or dimensionality reduction).
- Reinforcement Learning: The algorithm learns by interacting with an environment, using feedback from its own actions and experiences to make decisions.
- Semi-supervised Learning: This approach uses both labeled and unlabeled data for training – typically a small amount of labeled data with a large amount of unlabeled data.
- Deep Learning (see neural-networks): A subset of ML that uses neural networks with many layers (deep neural networks) to analyze various factors in complex tasks such as image recognition, natural language processing, and more.
Common Machine Learning Algorithms:
- Linear Regression
- Logistic Regression
- Decision Trees
- Random Forests
- Support Vector Machines (SVM)
- K-nearest Neighbors (KNN)
- neural-networks
Machine Learning Process:
- Data Collection: Gathering relevant and high-quality raw data.
- Data Preprocessing: Cleaning the data and transforming it into a suitable format for ML model training.
- Feature Selection/Engineering: Choosing the most relevant features or creating new features from existing ones.
- Model Choice: Selecting an appropriate algorithm for the problem at hand.
- Training: Feeding the preprocessed data into the model so it can learn from it.
- Evaluation: Testing the model against unseen data to assess its performance using metrics like accuracy, precision, recall, F1 score, etc.
- Hyperparameter Tuning/Optimization: Adjusting model parameters to improve performance.
- Deployment: Integrating the trained model into production environments where it can start making predictions or decisions on real-world data.
Applications of Machine Learning:
- Image recognition and computer vision
- Speech recognition
- Natural language processing
- Predictive analytics
- Autonomous vehicles
- Personalized recommendations (e.g., Netflix movie suggestions)
Challenges in Machine Learning:
- Overfitting: When a model performs well on training data but poorly on new, unseen data due to excessive complexity.
- Underfitting: When a model is too simple and cannot capture underlying trends in the training dataset.
- Data Privacy: Ensuring sensitive information used for training models remains confidential and secure.
- Bias: Avoiding unfair biases which can be present in training datasets and thus reflected in predictions made by ML models.
Machine learning in 2026
-
Modern generative AI is deep learning at scale: transformer-based large-language-models (see transformer-models-2023) are pre-trained with self-supervised learning on large corpora, then adapted with supervised fine-tuning and reinforcement learning from feedback. Open-weight models are covered in open-source-ai.
-
Classical ML (trees, forests, SVMs, linear models) is commonly still used for tabular data (general observation, not separately sourced). Reference libraries: scikit-learn (latest on PyPI at 2026-10-02: 1.9.1) and PyTorch (latest on PyPI at 2026-10-02: 2.14.1).
-
Typical challenges beyond those above: distribution shift, evaluation of generative systems, and regulation (see ai-regulation-and-policy).
Sources
- https://pypi.org/pypi/scikit-learn/json (fetched 2026-10-02)
- https://pypi.org/pypi/torch/json (fetched 2026-10-02)