Research map / Label-efficient and robust learning / Recognition and long-tail
Revisiting Bilinear Pooling: A Coding Perspective
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 |
|---|---|---|---|
| replace similarity-based bilinear coding with learned factorized sparse coding (FBC) | Bilinear pooling · VQA 2.0 val | accuracy (%): 59.8 -> 62.0 | Helped |
| FBC instead of explicit bilinear pooling | B-CNN · MINC | Average Precision (%): 74.5 -> 80.2 | Helped |
| add attention (2 glimpses) to FBC (ablation) | FBC · VQA 2.0 val | accuracy (%): 62.0 -> 65.7 | Helped |
| add attention (8 glimpses) to FBC (ablation) | FBC · VQA 2.0 test-dev | overall accuracy (%): 65.9 -> 69.7 | Helped |
| add counter model to FBC+Att (ablation) | FBC+Att · VQA 2.0 test-dev | overall accuracy (%): 69.7 -> 70.0 | Helped |
| increase number of atoms from 128 to 8192 (ablation) | FBC · MINC | accuracy (%): 79.3 -> 80.2 | Helped |
| increase atom rank r beyond small values (ablation) | FBC · MINC | accuracy (%) | Mixed |
| increase atoms from 512 to 2048 in FBC (ablation) | FBC · DTD | Average Precision (%): 70.3 -> 71.7 | Helped |
| increase atoms from 2048 to 8192 in FBC (ablation) | FBC · CUB | Average Precision (%): 83.6 -> 84.3 | No clear effect |
| FBC vs SMSO on CUB | FBC · CUB | Average Precision (%): 85.0 -> 84.3 | Hurt |
Closest papers by meaning
- Low-Rank Bilinear Pooling for Fine-Grained Classification (CVPR 2017 · similarity 0.90)
- Compact Bilinear Pooling (CVPR 2016 · similarity 0.90)
- Hadamard Product for Low-rank Bilinear Pooling (ICLR 2017 · similarity 0.89)
- Efficient Compact Bilinear Pooling via Kronecker Product (AAAI 2022 · similarity 0.87)
- SkipPool: Improved Sparse Hierarchical Graph Pooling with Differentiable Exploration (AAAI 2025 · similarity 0.82)
- The Balanced-Pairwise-Affinities Feature Transform (ICML 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.