Research map / Optimization / Optimizers
Deep Latent Dirichlet Allocation with Topic-Layer-Adaptive Stochastic Gradient Riemannian MCMC
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 |
|---|---|---|---|
| add topic- and layer-adaptive learning rates rather than fixed learning rates (ablation) | DLDA-TLASGR · 20News | per-heldout-word perplexity: 760 -> 757 | Helped |
| add topic- and layer-adaptive learning rates rather than fixed learning rates (ablation) | DLDA-TLASGR · RCV1 | per-heldout-word perplexity: 817 -> 815 | Helped |
| add topic- and layer-adaptive learning rates rather than fixed learning rates (ablation) | DLDA-TLASGR · Wiki | per-heldout-word perplexity: 789 -> 786 | Helped |
| use reduced-mean simplex parameters and avoid the expanded-mean pseudolikelihood assumption (ablation) | DLDA inference · 20News | per-heldout-word perplexity: 775 -> 757 | Helped |
| use reduced-mean simplex parameters and avoid the expanded-mean pseudolikelihood assumption (ablation) | DLDA inference · RCV1 | per-heldout-word perplexity: 827 -> 815 | Helped |
| use reduced-mean simplex parameters and avoid the expanded-mean pseudolikelihood assumption (ablation) | DLDA inference · Wiki | per-heldout-word perplexity: 792 -> 786 | Helped |
| increase model depth from one layer to three layers | DLDA-TLASGR · 20News | per-heldout-word perplexity: 770 -> 757 | Helped |
| increase model depth from one layer to three layers | DLDA-TLASGR · RCV1 | per-heldout-word perplexity: 823 -> 815 | Helped |
| increase model depth from one layer to three layers | DLDA-TLASGR · Wiki | per-heldout-word perplexity: 802 -> 786 | Helped |
| replace batch Gibbs sampling with mini-batch TLASGR inference | DLDA 128-64-32 · 20News | per-heldout-word perplexity: 752 -> 757 | Hurt |
| replace batch Gibbs sampling with mini-batch TLASGR inference | DLDA 128-64-32 · RCV1 | per-heldout-word perplexity: 802 -> 815 | Hurt |
Closest papers by meaning
- Poisson-Randomised DirBN: Large Mutation is Needed in Dirichlet Belief Networks (ICML 2021 · similarity 0.86)
- Deep Relational Topic Modeling via Graph Poisson Gamma Belief Network (NeurIPS 2020 · similarity 0.86)
- Convolutional Poisson Gamma Belief Network (ICML 2019 · similarity 0.86)
- Deep Poisson gamma dynamical systems (NeurIPS 2018 · similarity 0.85)
- Discriminative Topic Modeling with Logistic LDA (NeurIPS 2019 · similarity 0.85)
- Investigating the Overlooked Hessian Structure: From CNNs to LLMs (ICML 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.