Research map / Label-efficient and robust learning / Recognition and long-tail
Inter-battery Topic Representation Learning
Cheng Zhang, Hedvig Kjellström, Carl Henrik Ek
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 |
|---|---|---|---|
| use IBTM shared/private factorized topics instead of plain LDA with same number of topics | topic model on Conv5_1 features · LabelMe (image-image) | accuracy: 85.25% -> 89.75% | Improved |
| use IBTM 15 shared topics instead of full 1024-dim SVM | classifier on Conv5_1 features · LabelMe (image-image) | accuracy: 87% -> 89.75% | Improved |
| use LDA topics instead of PCA with same dimensions | 15-dim representation · LabelMe (image-image) | accuracy: 80.88% -> 85.25% | Improved |
| use IBTM instead of SWB (text-specific factorization) | factorized topic model · LabelMe (image-image) | accuracy: 59.88% -> 89.75% | Improved |
| add the annotation view at test time (2 views instead of image only) | IBTM15 · LabelMe (image-annotation) | accuracy: 89.38% -> 95% | Improved |
| train IBTM with annotations, test with images only, versus full SVM | IBTM15 1V · LabelMe (image-annotation) | accuracy: 87.63% -> 89.38% | Improved |
| use IBTM image-image instead of full SVM | classifier · Leeds Butterfly (image-image) | accuracy: 95.49% -> 95.86% | Improved |
| use IBTM instead of LDA15 | topic representation · Leeds Butterfly (image-image) | accuracy: 91.92% -> 95.86% | Improved |
| add segmentation view at test time | IS IBTM15 · Leeds Butterfly (image-segmentation) | accuracy: 96.05% -> 99.06% | Improved |
| use image+segmentation concatenated SWB instead of image-only SWB | SWB15 · Leeds Butterfly | accuracy: 80.26% -> 94.55% | Improved |
Closest papers by meaning
- Sawtooth Factorial Topic Embeddings Guided Gamma Belief Network (ICML 2021 · similarity 0.85)
- Inter and Intra Topic Structure Learning with Word Embeddings (ICML 2018 · similarity 0.85)
- Bayesian Deep Embedding Topic Meta-Learner (ICML 2022 · similarity 0.85)
- TopicNet: Semantic Graph-Guided Topic Discovery (NeurIPS 2021 · similarity 0.84)
- Topic Modeling as Multi-Objective Contrastive Optimization (ICLR 2024 · similarity 0.84)
- Topic-VQ-VAE: Leveraging Latent Codebooks for Flexible Topic-Guided Document Generation (AAAI 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.
