Research map / Vision-language and multimodal / CLIP and zero-shot
Subspace Alignment for CLIP-based Continual Learning via Canonical Correlation Analysis
Huan Zhang, Shuyu Dong, Yujin Zheng, Dingwen Wang, Shenghua Fan, Fan Lyu
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 CCA subspace projection (ablation) | Continual-CLIP · CIFAR-100 | last accuracy: 66.7 -> 76.5 | Improved |
| add Random Fourier Projection to CCA (ablation) | CCA-CL (no FTT) · CIFAR-100 | last accuracy: 76.5 -> 81.0 | Improved |
| add first-task fine-tuning (ablation) | CCA-CL · CIFAR-100 | last accuracy: 81.0 -> 83.0 | Improved |
| lower energy threshold from 0.99 to 0.9 (ablation) | CCA-CL (no FTT) · CIFAR-100 | last accuracy: 81.0 -> 76.5 | Hurt |
| raise energy threshold from 0.99 to 1 (ablation) | CCA-CL (no FTT) · CIFAR-100 | last accuracy: 81.0 -> 80.8 | No clear effect |
| widen RFP from 6144 to 10240 (ablation) | CCA-CL (no FTT) · CIFAR-100 | last accuracy: 81.0 -> 80.2; memory 6.3 -> 11.9 G | Hurt |
| narrow RFP from 6144 to 2048 (ablation) | CCA-CL (no FTT) · CIFAR-100 | last accuracy: 81.0 -> 79.7 | Hurt |
| use CCA-CL instead of RAPF for efficiency | CIFAR-100 training · CIFAR-100 | learning time (minutes): 13.9 -> 5.8 | Improved |
| remove first-task training (ablation) | CCA-CL · CIFAR-100 | learning time (minutes): 5.8 -> 3.9 | 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.
- CLAP — CLAP: Isolating Content from Style through Contrastive Learning with Augmented Prompts (ECCV 2024) · better on 1 of 1 shared result
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
- Embracing Language Inclusivity and Diversity in CLIP through Continual Language Learning (AAAI 2024 · similarity 0.86)
- Refining CLIP's Spatial Awareness: A Visual-Centric Perspective (ICLR 2025 · similarity 0.86)
- Post-pre-training for Modality Alignment in Vision-Language Foundation Models (CVPR 2025 · similarity 0.85)
- CLAP4CLIP: Continual Learning with Probabilistic Finetuning for Vision-Language Models (NeurIPS 2024 · similarity 0.85)
- SmartCLIP: Modular Vision-language Alignment with Identification Guarantees (CVPR 2025 · similarity 0.85)
- BiLoRA: Almost-Orthogonal Parameter Spaces for Continual Learning (CVPR 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.
