Neblux Knowledge Graph
Deep Learning
Deep learning refers to a class of machine learning methods that use artificial neural networks with many layers to learn hierarchical representations from large datasets.
Overview
The foundation of deep learning is the artificial neural network, loosely inspired by biological neural circuits. Multi-layer networks with nonlinear activation functions can approximate arbitrarily complex functions. The key enabling breakthrough was the development of efficient backpropagation training in the 1980s, later combined with large datasets and GPU computing in the 2010s to produce dramatic advances. Convolutional neural networks (CNNs) transformed image recognition; recurrent networks advanced sequence modeling; and transformer architectures, introduced in 2017, enabled large-scale language models. The ImageNet competition results in 2012 marked the beginning of widespread industrial adoption.
Why it matters
Deep learning has fundamentally changed artificial intelligence and influenced virtually every domain it has touched. It achieved human-level or superhuman performance in image classification, game playing, protein structure prediction, and speech recognition. Language models built on deep learning have transformed natural language processing and generated new tools for science, medicine, and engineering. The field has also raised important questions in philosophy of mind, ethics, and cognitive science about the nature of intelligence. Its rapid advance shaped technology policy and created major new economic sectors in the 21st century.
What it builds on
Related concepts
- Computer ScienceappliedDeep learning has become the dominant paradigm in computer science for tasks including image recognition, speech recognition, and language modeling.
- NeuroscienceconceptualDeep neural networks draw conceptual inspiration from the hierarchical processing of information in biological neural circuits in the brain.
- Natural Language ProcessingappliedTransformer-based deep learning models have revolutionized natural language processing by achieving state-of-the-art results across language benchmarks.
- Probability TheorylogicalProbabilistic loss functions, Bayesian interpretations, and stochastic training procedures ground deep learning in probability theory.