AI Science Tracker
EN

Comparing Classification and Feature Learning by Analyzing Their Loss Functions in Normalized Spaces with Different Scale Factors

Shenzhen University; University of Oxford; North Minzu University · 5 Nov 2026 · computer science

Discovery

Comparing Classification and Feature Learning by Analyzing Their Loss Functions in Normalized Spaces with Different Scale Factors

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The paper analyzes how cross-entropy and binary cross-entropy losses affect classification and feature geometry in normalized spaces, then tests the analysis with neural-network experiments. Under the tested training setup, scale factors changed classification accuracy and feature quality differently; the conclusions are tied to the reported losses and experiments.

Technology: Controlled neural-network training and normalized feature-space analysis

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The paper analytically compares normalized cross-entropy and binary cross-entropy objectives and tests scale-factor effects on classifier training and learned feature geometry.
Limitation / still unknown
The study evaluates the reported models and datasets only; broader generalization and real-world deployment outcomes were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07741v1 · https://arxiv.org/html/2611.07741v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open in the full catalog for filters, charts and related discoveries. Browse computer science discoveries.