Research map / Label-efficient and robust learning / Self-supervised and contrastive
Target-Embedding Autoencoders for Supervised Representation Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add target reconstruction and learn a compact target embedding | Linear prediction model · UKCF infections | Prc: 0.351 ± 0.079 -> 0.450 ± 0.035 | Helped |
| add target reconstruction and learn a compact target embedding | GRU prediction model · ADNI biomarkers | Mse: 0.092 ± 0.012 -> 0.063 ± 0.010 | Helped |
| add feature reconstruction alongside target reconstruction (ablation) | Tea · UKCF infections | Prc: 0.450 ± 0.035 -> 0.414 ± 0.028 | Hurt |
| omit final joint training (ablation) | Linear Tea · UKCF infections | Prc: 0.402 ± 0.026 vs 0.450 ± 0.035 | Hurt |
| skip the initial staged training (ablation) | Linear Tea · UKCF infections | Prc: 0.431 ± 0.031 vs 0.450 ± 0.035 | Hurt |
| replace direct prediction loss with latent loss in joint training (ablation) | Linear Tea · UKCF infections | Prc: 0.435 ± 0.031 vs 0.450 ± 0.035 | No clear effect |
| include latent and prediction losses with reconstruction (ablation) | Linear Tea · UKCF infections | Prc: 0.454 ± 0.036 vs 0.450 ± 0.035 | No clear effect |
| replace direct prediction loss with latent loss in joint training (ablation) | GRU Tea · ADNI biomarkers | Mse: 0.058 ± 0.012 vs 0.063 ± 0.010 | No clear effect |
| increase the reconstruction-loss weight to the extreme λ=1 (ablation) | Tea · UKCF | Roc: Roc ∼ 0.5 | Hurt |
| combine target reconstruction with feature reconstruction (ablation) | Tea · UKCF infections | Prc: 0.450 ± 0.035 -> 0.414 ± 0.028 | Hurt |
Closest papers by meaning
- Scaling and evaluating sparse autoencoders (ICLR 2025 · similarity 0.85)
- Enhancing Evolving Domain Generalization through Dynamic Latent Representations (AAAI 2024 · similarity 0.84)
- Exploring Target Representations for Masked Autoencoders (ICLR 2024 · similarity 0.84)
- Mechanistic Interpretability with Sparse Autoencoder Neural Operators (NeurIPS 2026 · similarity 0.84)
- Projecting Assumptions: The Duality Between Sparse Autoencoders and Concept Geometry (NeurIPS 2025 · similarity 0.83)
- Joint‑Embedding vs Reconstruction: Provable Benefits of Latent Space Prediction for Self‑Supervised Learning (NeurIPS 2025 · similarity 0.83)
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.