Research map / Optimization / Optimizers
Scalable Planning with Tensorflow for Hybrid Nonlinear Domains
Ga Wu, Buser Say, Scott Sanner
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 Tensorflow backpropagation planning with RMSProp instead of fmincon | Hybrid nonlinear planning · Nonlinear domains: Navigation, Reservoir Control, and HVAC | plan cost | Improved |
| Use Tensorflow backpropagation planning with RMSProp instead of a domain-specific heuristic | Hybrid planning · Linear and nonlinear benchmark domains | plan cost | Improved |
| Use Tensorflow with RMSProp rather than MILP optimization | Hybrid planning · Linear domains | plan cost | Mixed |
| Use Tensorflow with RMSProp instead of fmincon | Nonlinear Reservoir Control planning · Reservoir Control, horizon 60 | optimization speed: 16 times faster | Improved |
| Use Tensorflow with RMSProp instead of fmincon | Nonlinear Reservoir Control planning · Reservoir Control, horizon 120 | optimization speed: hundreds of times faster | Improved |
| Use RMSProp rather than the other four tested Tensorflow optimizers | Tensorflow planning · HVAC | convergence speed | Improved |
| Use Adagrad or Adadelta rather than RMSProp | Tensorflow planning · HVAC | optimization progress | Hurt |
| Set the learning rate to 0.1 instead of 0.001 (ablation) | Tensorflow with RMSProp · HVAC | initial convergence speed | Improved |
| Set the learning rate to 0.1 instead of 0.001 (ablation) | Tensorflow with RMSProp · HVAC | best final score reached | Hurt |
| Set the learning rate to 0.01 (ablation) | Tensorflow with RMSProp · HVAC | convergence to near-optimal score | Improved |
| Use a larger learning rate rather than a smaller one (ablation) | Tensorflow with RMSProp · HVAC | initial convergence speed | Mixed |
| Scale nonlinear planning to 576,000 continuous action parameters across 100 parallel instances | Tensorflow with RMSProp · HVAC nonlinear, horizon 96 | convergence time: 576,000 actions; 4 minutes | Improved |
Closest papers by meaning
- Deep Reactive Policies for Planning in Stochastic Nonlinear Domains (AAAI 2019 · similarity 0.86)
- Generalized Policy Iteration using Tensor Approximation for Hybrid Control (ICLR 2024 · similarity 0.85)
- Tensor Program Optimization with Probabilistic Programs (NeurIPS 2022 · similarity 0.85)
- RoME: Domain-Robust Mixture-of-Experts for MILP Solution Prediction across Domains (NeurIPS 2025 · similarity 0.84)
- Synergistic Tensor and Pipeline Parallelism (NeurIPS 2025 · similarity 0.84)
- B$^3$-PWL: GPU-Batched Branch-and-Bound for Piecewise-Linear Optimization with SOS2 Constraints (NeurIPS 2026 · 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.
