Research map / Probabilistic and scientific modeling / Variational and generative sampling
Variational Inference for Cyclic Learning
Zhuojun Zou, Jie Hao
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 |
|---|---|---|---|
| replace GAN discriminator with EM cyclic training (CycleGN) | CycleGAN · Cityscapes labels->photo | FCN per-pixel accuracy: 0.52 -> 0.52 | No clear effect |
| use CycleGAN instead of CycleGN | CycleGN · Cityscapes photo->labels | FCN per-pixel accuracy: 0.51 -> 0.58 | Improved |
| add EM training (CycleTrack-EM) over single-step CycleTrack | CycleTrack · LaSOT (unsupervised, detector-annotated) | AUC: 51.0 -> 56.5 | Improved |
| add EM training over single-step CycleTrack | CycleTrack · TrackingNet (unsupervised, detector-annotated) | AUC: 75.9 -> 77.3 | Improved |
| add EM training over single-step CycleTrack | CycleTrack · LaSOT (strictly unsupervised) | AUC: 45.0 -> 51.2 | Improved |
| add EM training over single-step CycleTrack | CycleTrack · TrackingNet (strictly unsupervised) | AUC: 65.6 -> 69.1 | Improved |
| add the E-step (forward-process freezing) to cycle-consistency-only training (ablation) | CycleTrack-EM · LaSOT | tracking quality | Improved |
| add alternating EM steps instead of rotated per-batch objectives (ablation) | CycleGN · Cityscapes labels->photo | FCN per-pixel accuracy: 0.51 -> 0.52 | No clear effect |
| optimize only one direction (Dx only) instead of joint (ablation) | CycleGN · Cityscapes labels->photo | FCN per-pixel accuracy: 0.41 -> 0.52 | Hurt |
| use full supervision instead of semi-supervision | CycleTrack-EM · LaSOT | AUC: 63.6 -> 70.1 | Mixed |
Closest papers by meaning
- Augmented CycleGAN: Learning Many-to-Many Mappings from Unpaired Data (ICML 2018 · similarity 0.85)
- A Condition Number for Joint Optimization of Cycle-Consistent Networks (NeurIPS 2019 · similarity 0.85)
- $C^2M^3$: Cycle-Consistent Multi-Model Merging (NeurIPS 2024 · similarity 0.84)
- LoRanPAC: Low-rank Random Features and Pre-trained Models for Bridging Theory and Practice in Continual Learning (ICLR 2025 · similarity 0.84)
- VCT: Training Consistency Models with Variational Noise Coupling (ICML 2025 · similarity 0.84)
- Convergence and Implicit Bias of Gradient Descent on Continual Linear Classification (ICLR 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.
