Glossary Library

Technical GlossaryDeep Learning

Masked Language Modeling

A pretraining objective based on masking some input tokens and predicting them from context.

Masked Language Modeling plays a central role in the pretraining of encoder-based language models. The model learns to predict hidden tokens from the surrounding context. This supports not just generation, but deeper contextual understanding. It is a powerful and historically influential pretraining paradigm in modern representation learning.

You Might Also Like

Explore these concepts to continue your artificial intelligence journey.

Glossary Cover

yapay-zeka-temelleri

Representation Learning

An approach in which informative, discriminative, and task-relevant internal representations are learned automatically from raw data.

Glossary Cover

makine-ogrenmesi

Non-negative Matrix Factorization

A dimensionality reduction technique that produces part-based and interpretable representations in non-negative data.

Glossary Cover

makine-ogrenmesi

Autoencoder-Based Dimensionality Reduction

An approach that learns lower-dimensional representations of data through neural-network-based compression.