Machine learning · TechGene

Definition

Methods that adjust a model using data or experience from interaction to improve a system's performance on a target task.

Samuel's checkers research, published in 1959, demonstrated a program improving its performance through experience. Machine learning later developed a variety of models and learning settings. This entry does not treat that paper as the common origin of all learning algorithms.

Learning outcomes depend on data, objective functions, and evaluation methods. Improvements on the training set must be tested on unseen data or real tasks and cannot be taken directly as evidence of generalization.

Principles and impact

Problems solved

Reduced dependence on manually writing individual decision rules.

Principles

Adjusts model parameters or structure using experience and feedback on objectives.

Improvements

Not recorded

Significance

Supported a broad range of applications, including recognition, prediction, generation, and decision-making.

History and applications

People

Arthur L. Samuel

Organizations

Not recorded

Events

Not recorded

Applications

Classification、Prediction、Pattern recognition

Limitations

Data bias, distribution shifts, and evaluation leakage can cause apparent results to diverge from actual capabilities.

Upstream and downstream relations

Upstream

←Artificial neural networksusesArtificial neural networks with trainable parameters can serve as learning models; early fixed logical networks did not automatically have the ability to learn.

Sources: A Logical Calculus of the Ideas Immanent in Nervous Activity Learning Representations by Back-propagating Errors

←Artificial intelligencespecializesWithin the engineering scope of AI covered by this family, machine learning improves performance through data and experience; AI also includes approaches that do not center on learning.

Sources: Some Studies in Machine Learning Using the Game of Checkers What is Artificial Intelligence? — Basic Questions

←PerceptronusesThe perceptron adjusts its weights through feedback from examples, illustrating the application of neural models to learning classification tasks.

Sources: The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain Learning Representations by Back-propagating Errors

Downstream

→Deep learninghas specialized formDeep learning uses multiple layers of trainable representations to learn, making it a class of machine-learning methods.

Sources: A Fast Learning Algorithm for Deep Belief Nets ImageNet Classification with Deep Convolutional Neural Networks

→Diffusion generative modelshas specialized formDiffusion generative models learn a reverse process from data and form a class of generative machine-learning methods.

Sources: Deep Unsupervised Learning using Nonequilibrium Thermodynamics Denoising Diffusion Probabilistic Models

→Reinforcement learninghas specialized formReinforcement learning learns to make decisions from rewards received through interaction with an environment and is the branch of machine learning concerned with sequential decision-making.

Sources: Q-learning

→Support vector machineshas specialized formSupport vector machines are a supervised-learning method with an explicit margin criterion, representing a specific approach within machine learning.

Sources: Support-vector Networks

Definition

Principles and impact

Problems solved

Principles

Improvements

Significance

History and applications

People

Organizations

Events

Applications

Limitations

Upstream and downstream relations

Upstream

Downstream

Explore this technology's connections