Research map / Label-efficient and robust learning / Recognition and long-tail
Set Representation Auxiliary Learning with Adversarial Encoding Perturbation and Optimization
Yankai Chen, Xinni Zhang, Henry Peng Zou, Bowei He, Yangning Li, Philip S. Yu, Irwin King, Xue Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 Wasserstein-based SFE encoder instead of mean-pooling (ablation) | SRAL · Friendster | Recall@20: 67.02 -> 91.57 | Improved |
| add linear interpolation instead of two-layer MLP for dimension completion (ablation) | SRAL · Friendster | Recall@20: 75.45 -> 91.57 | Improved |
| add adversarial encoding perturbation optimization (AEPO) (ablation) | SRAL · Friendster | Recall@20: 77.13 -> 91.57 | Improved |
| add adversarial learning (ablation) | SRAL · Friendster | Recall@20: 87.38 -> 91.57 | Improved |
| add adversarial learning (ablation) | SRAL · Bundle recommendation Youshu | Recall@20: 26.94 -> 26.92 | No clear effect |
| add AEPO (ablation) | SRAL · LDA-3k | AUC: 66.21 -> 87.93 | Improved |
| add AEPO (ablation) | SRAL · ModelNet40 (MLP) | accuracy: 86.56 -> 86.53 | No clear effect |
| increase Monte Carlo trials R from 4 to 32 (ablation) | SRAL · Friendster | Recall@20: 41.23 -> 91.57 | Improved |
| use Sinkhorn/KL/JS distance instead of Wasserstein (ablation) | SRAL · Friendster | Recall@20 | Hurt |
| perturb intermediate encoding instead of final embeddings (DIJ) (ablation) | SRAL · Friendster | Recall@20 | Improved |
| swap SSL loss for triplet, SNN or Barlow Twins (ablation) | SRAL · Friendster | Recall@20 | No clear effect |
| add SRAL+ to CrossCBR | CrossCBR · Youshu | Recall@20: 26.41 -> 26.92 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FSPool — FSPool: Learning Set Representations with Featurewise Sort Pooling (ICLR 2020) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Differentiable Expectation-Maximization for Set Representation Learning (ICLR 2022 · similarity 0.86)
- Boosting Open Set Recognition Performance through Modulated Representation Learning (ICLR 2026 · similarity 0.85)
- Polynomial Width is Sufficient for Set Representation with High-dimensional Features (ICLR 2024 · similarity 0.84)
- Synergy Between Sufficient Changes and Sparse Mixing Procedure for Disentangled Representation Learning (ICLR 2025 · similarity 0.82)
- $\alpha$TC-VAE: On the relationship between Disentanglement and Diversity (ICLR 2024 · similarity 0.82)
- Graph-based Unsupervised Disentangled Representation Learning via Multimodal Large Language Models (NeurIPS 2024 · similarity 0.82)
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.
