Research map / Probabilistic and scientific modeling / Variational and generative sampling
Sequential Neural Models with Stochastic Layers
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 |
|---|---|---|---|
| Combine recurrent deterministic states with Markov stochastic latent states | SRNN · Blizzard and TIMIT speech test sets | average log-likelihood: Blizzard: ≥ 11991 vs ≈ 9516; TIMIT: ≥ 60550 vs ≈ 30235 | Helped |
| Combine recurrent deterministic states with Markov stochastic latent states | SRNN · Nottingham, JSB chorales, MuseData, and Piano-midi.de | average log-likelihood: ≥ -2.94 vs ≈ -2.31; ≥ -4.74 vs ≈ -3.99; ≥ -6.28 vs ≈ -5.60; ≥ -8.20 vs ≈ -7.05 | Mixed |
| Add smoothing with future sequence information (ablation) | SRNN · Blizzard test | average log-likelihood per 0.5s sequence: ≥ 10846 -> ≥ 10991; ≥ 10572 -> ≥ 11991 | Helped |
| Add smoothing with future sequence information (ablation) | SRNN · TIMIT test | average log-likelihood per sequence: ≥ 50524 -> ≥ 59269; ≥ 52126 -> ≥ 60550 | Helped |
| Add Res_q predictive-prior residual parameterization (ablation) | SRNN with smoothing · Blizzard test | average log-likelihood per 0.5s sequence: ≥ 10991 -> ≥ 11991 | Helped |
| Add Res_q predictive-prior residual parameterization (ablation) | SRNN with filtering · Blizzard test | average log-likelihood per 0.5s sequence: ≥ 10846 -> ≥ 10572 | Hurt |
| Add Res_q predictive-prior residual parameterization (ablation) | SRNN with smoothing · TIMIT test | average log-likelihood per sequence: ≥ 59269 -> ≥ 60550 | Helped |
| Add Res_q predictive-prior residual parameterization (ablation) | SRNN with filtering · TIMIT test | average log-likelihood per sequence: ≥ 50524 -> ≥ 52126 | Helped |
| Use Res_q and smoothing rather than direct mean parameterization and filtering (ablation) | SRNN · Polyphonic music datasets | average log-likelihood | No clear effect |
Closest papers by meaning
- Z-Forcing: Training Stochastic Recurrent Networks (NeurIPS 2017 · similarity 0.90)
- Inferring stochastic low-rank recurrent neural networks from neural data (NeurIPS 2024 · similarity 0.88)
- Variational Smoothing in Recurrent Neural Network Language Models (ICLR 2019 · similarity 0.87)
- BARNN: A Bayesian Autoregressive and Recurrent Neural Network (ICML 2025 · similarity 0.87)
- Continuous Autoregressive Modeling with Stochastic Monotonic Alignment for Speech Synthesis (ICLR 2025 · similarity 0.85)
- Latent Optimal Paths by Gumbel Propagation for Variational Bayesian Dynamic Programming (ICML 2024 · 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.