Research map / Vision-language and multimodal / Human and embodied modeling
A Deep and Autoregressive Approach for Topic Modeling of Multimodal Data
Yin Zheng, Yu‐Jin Zhang, Hugo Larochelle
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 class-label supervision to DocNADE (SupDocNADE) | DocNADE · LabelMe | accuracy (%): 81.97 -> 83.43 | Improved |
| add class-label supervision to DocNADE (SupDocNADE) | DocNADE · UIUC-Sports | accuracy (%): 74.23 -> 77.29 | Improved |
| add class-label supervision to DocNADE (SupDocNADE) | DocNADE · LabelMe | annotation F-measure (%): 43.32 -> 43.87 | No clear effect |
| add class-label supervision to DocNADE (SupDocNADE) | DocNADE · UIUC-Sports | annotation F-measure (%): 46.38 -> 46.95 | No clear effect |
| SupDocNADE instead of sLDA for annotation | sLDA · LabelMe | annotation F-measure (%): 38.7 -> 43.87 | Improved |
| SupDocNADE instead of sLDA for annotation | sLDA · UIUC-Sports | annotation F-measure (%): 35.0 -> 46.95 | Improved |
| use SupDocNADE instead of MMLDA for annotation on LabelMe | MMLDA · LabelMe | annotation F-measure (%): 46.64 -> 43.87 | Hurt |
| add spatial 2x2 grid and tuned generative weight lambda (ablation) | SupDocNADE · LabelMe and UIUC-Sports | accuracy | Mixed |
| add more hidden layers (2 instead of 1), 625 epochs pretraining (ablation) | SupDeepDocNADE · MIR Flickr | MAP: 0.654 -> 0.671 | Improved |
| add third hidden layer, 625 epochs pretraining (ablation) | SupDeepDocNADE · MIR Flickr | MAP: 0.671 -> 0.670 | No clear effect |
| add third hidden layer, 4125 epochs pretraining (ablation) | SupDeepDocNADE · MIR Flickr | MAP: 0.684 -> 0.691 | Improved |
| increase unsupervised pretraining from 625 to 4125 epochs (3 layers) (ablation) | SupDeepDocNADE · MIR Flickr | MAP: 0.670 -> 0.691 | Improved |
Closest papers by meaning
- Document Informed Neural Autoregressive Topic Models with Distributional Prior (AAAI 2019 · similarity 0.89)
- textTOvec: DEEP CONTEXTUALIZED NEURAL AUTOREGRESSIVE TOPIC MODELS OF LANGUAGE WITH DISTRIBUTED COMPOSITIONAL PRIOR (ICLR 2019 · similarity 0.87)
- Deep Autoencoding Topic Model With Scalable Hybrid Bayesian Inference (TPAMI 2020 · similarity 0.87)
- Sparse Autoencoders are Topic Models (ICML 2026 · similarity 0.87)
- Document Neural Autoregressive Distribution Estimation (JMLR 2017 · similarity 0.87)
- Topic-VQ-VAE: Leveraging Latent Codebooks for Flexible Topic-Guided Document Generation (AAAI 2024 · similarity 0.84)
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.
