representations

Definition

The internal numerical forms — vectors, feature maps, embeddings — into which a model maps its input. A good representation places semantically similar inputs near one another, so that downstream tasks become solvable with a simple classifier on top. Representation learning is defined by this goal: the objective is the quality of the internal encoding itself, not performance on any one end task.

Explainers using this term

• Paper Walkthrough: Code as Worlds — An Agent That Writes the World Down as Runnable CodeCode as Worlds: Agentic Discovery of Executable World Representations for Physical Reasoning
• Self-Supervised Learning — The Day Unlabeled Data Became an AssetA Simple Framework for Contrastive Learning of Visual Representations
• Embeddings from Scratch — from word2vec Intuition to Contextual EmbeddingsEfficient Estimation of Word Representations in Vector Space