Research map / Label-efficient and robust learning / Recognition and long-tail
How Classifier Features Transfer to Downstream: An Asymptotic Analysis in a Two-Layer Model
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Theoretical asymptotic analysis of when classifier-learned features transfer to unseen classes via clustering, in a two-layer network model.
Problem
Feature transfer from networks trained for classification to new tasks is known to benefit from larger datasets, but the theoretical conditions under which transferred features cluster unseen classes effectively remain unclear.
Approach
Analyzes feature transfer in a two-layer network trained for classification, asymptotically characterizing conditions under which features enable effective clustering of unseen classes.
Closest papers by meaning
- Why Do Better Loss Functions Lead to Less Transferable Features? (NeurIPS 2021 · similarity 0.90)
- Transfer Learning in Infinite Width Feature Learning Networks (ICLR 2026 · similarity 0.88)
- Features are fate: a theory of transfer learning in high-dimensional regression (ICML 2025 · similarity 0.88)
- Going Beyond Neural Network Feature Similarity: The Network Feature Complexity and Its Interpretation Using Category Theory (ICLR 2024 · similarity 0.87)
- What is being transferred in transfer learning? (NeurIPS 2020 · similarity 0.87)
- From Kernels to Features: A Multi-Scale Adaptive Theory of Feature Learning (ICML 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.