Research map / Optimization / Optimizers
NOVAS: Non-convex Optimization via Adaptive Stochastic Search for End-to-end Learning and Control
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 NOVAS instead of unrolled DCEM | SPEN regression · SPEN regression | speed: 5x | Helped |
| Use NOVAS instead of unrolled DCEM | SPEN regression · SPEN regression | inner-loop convergence rate | Helped |
| Use sampling-based optimization (NOVAS or unrolled DCEM) instead of unrolled gradient descent | SPEN energy learning · SPEN regression | energy landscape quality | Helped |
| Change inner-loop optimization hyperparameters at inference after training with unrolled gradient descent | SPEN energy learning · SPEN regression | regression performance | Hurt |
| Replace the explicit Hamiltonian minimizer with NOVAS | Deep FBSDE controller · Cart-pole swing-up | control performance | No clear effect |
| Use NOVAS-FBSDE rather than constant equal allocation | Portfolio strategy · 101-dimensional portfolio optimization | probability of outperforming the market: almost 5% | Helped |
| Use NOVAS-FBSDE rather than random allocations | Portfolio strategy · 101-dimensional portfolio optimization | probability of outperforming the market: almost 5% | Helped |
| Avoid unrolling the NOVAS inner-loop graph (ablation) | NOVAS · SPEN regression | speed: 5x | Helped |
| Unroll NOVAS rather than keep its inner loop off-graph (ablation) | NOVAS · SPEN regression | learning performance | No clear effect |
| Use the sigmoid-like shape function rather than the exponential shape function (ablation) | NOVAS · SPEN regression | accuracy | Helped |
| Use the exponential shape function rather than the sigmoid-like shape function (ablation) | NOVAS · SPEN regression | speed | Helped |
| Use either tested shape function instead of the other (ablation) | NOVAS-FBSDE · Control using FBSDEs | performance | No clear effect |
Closest papers by meaning
- Adaptive Stochastic Natural Gradient Method for One-Shot Neural Architecture Search (ICML 2019 · similarity 0.88)
- End-to-End Learning for Stochastic Optimization: A Bayesian Perspective (ICML 2023 · similarity 0.87)
- Towards Better Understanding of Adaptive Gradient Algorithms in Generative Adversarial Nets (ICLR 2020 · similarity 0.87)
- Adaptive Methods through the Lens of SDEs: Theoretical Insights on the Role of Noise (ICLR 2025 · similarity 0.86)
- BPQP: A Differentiable Convex Optimization Framework for Efficient End-to-End Learning (NeurIPS 2024 · similarity 0.86)
- Non-asymptotic Analysis of Biased Adaptive Stochastic Approximation (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.