Article

Scaffolded representation learning in deep networks

2026-04-16

Abstract excerpt

<title>Abstract</title> <p>Deep networks learn coarse structure before fine-grained distinctions, yet whether coarse structure actively scaffolds later differentiation remains untested. Here we show that representations assemble through a load-bearing scaffold. Tracking features at per-sample resolution across 55 runs, three architecture families and two training datasets, we find a reproducible three-phase progr...

Topics

Open a Topic to create a Post that cites this publication.

Identifiers and source

Literature Corpus work: d4fe1552-7552-503c-9a36-8f007494c332
DOI: 10.21203/rs.3.rs-9269961/v1

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

Scaffolded representation learning in deep networksDOI 10.21203/rs.3.rs-9269961/v1Read

Learning sculpts orthogonal task manifolds for continual skill learning in recurrent networksSemantic proximity Read

An invariant schema emerges within a neural network during hierarchical learning of visual boundariesSemantic proximity Read

Cluster-Based Inference for Memory-Based CognitionSemantic proximity Read

Hierarchical Emergence Profiles of Human-Derived Dimensions are a Fundamental Property of Deep Neural NetworksSemantic proximity Read

Backpropagation-Free Continual Learning: A Unified Energy-Based Framework, Reproducible Benchmark, and Open ChallengesSemantic proximity Read

Leveraging prior concept learning improves ability to generalize from few examples in computational models of human object recognitionSemantic proximity Read

A biologically inspired architecture with switching units can learn to generalize across backgroundsSemantic proximity Read

Abstract representations emerge naturally in neural networks trained to perform multiple tasksSemantic proximity Read

Gradual sensory maturation promotes abstract representation learningSemantic proximity Read

A controller-peripheral architecture and costly energy principle for learningSemantic proximity Read

Why Compression Creates Intelligence: The Architecture of Experience in Large ModelsSemantic proximity Read

Neural representations supporting generalization under continual learningSemantic proximity Read