Research map / Probabilistic and scientific modeling / Dynamical systems
Collapsed Amortized Variational Inference for Switching Nonlinear Dynamical Systems
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 |
|---|---|---|---|
| Use nonlinear continuous dynamics instead of linear dynamics | SNLDS versus SLDS · Dubins path | frame-wise F1: 61.5 ± 8.0 -> 84.3 ± 7.2 | Helped |
| Replace exact discrete-state marginalization with Gumbel-Softmax relaxation (ablation) | SNLDS · Reacher Task | switching-point F1: 78.1 ± 4.2 -> 5.0 ± 8.7 | Hurt |
| Replace exact discrete-state marginalization with Gumbel-Softmax relaxation (ablation) | SNLDS · Bouncing Ball | frame-wise F1: 100. -> 93.8 ± 4.0 | Hurt |
| Add entropy regularization that encourages uniform use of discrete states (ablation) | SNLDS · Reacher Task | segmentation F1 | Helped |
| Delay entropy-coefficient and temperature annealing until later in training (ablation) | SNLDS · Dubins path | frame-wise F1 | Helped |
| Train without entropy regularization (ablation) | SNLDS · Dubins path | likelihood | Hurt |
| Use nonlinear continuous dynamics rather than linear transitions | SNLDS versus SLDS · Reacher Task | switching-point F1: 59.6 ± 3.2 -> 78.1 ± 4.2 | Helped |
Closest papers by meaning
- Variational Inference for Continuous-Time Switching Dynamical Systems (NeurIPS 2021 · similarity 0.89)
- End-to-End Identifiable and Consistent Recurrent Switching Dynamical Systems (NeurIPS 2026 · similarity 0.86)
- Amortized Control of Continuous State Space Feynman-Kac Model for Irregular Time Series (ICLR 2025 · similarity 0.86)
- Identifying latent state transitions in non-linear dynamical systems (ICLR 2025 · similarity 0.85)
- Closing the Approximation Gap in Simulation-free Latent SDEs (NeurIPS 2026 · similarity 0.85)
- Gated Inference Network: Inference and Learning State-Space Models (NeurIPS 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.