Research map / Optimization / Nonconvex and smooth optimization
Sensitivity-Free Gradient Descent Algorithms
Ion Matei, Maksym Zhenirovskyy, Johan de Kleer, John Maxwell
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 |
|---|---|---|---|
| replace sensitivity-enabled ODE solver with Algorithm 2 (no ODE solver, block coordinate descent) | Algorithm 1 direct, Pytorch · Cucker-Smale N=5 | time per epoch (s): 0.463 -> 0.009 (50x) | Improved |
| replace Algorithm 1 with Algorithm 2 | Algorithm 1 direct, Pytorch · Cucker-Smale N=200 | time per epoch (s): 0.614 -> 0.155 (4x) | Improved |
| replace Algorithm 1 with Algorithm 3 (ODE solve, no sensitivities) | Algorithm 1 direct, Pytorch · Cucker-Smale N=5 | time per epoch (s): 0.463 -> 0.171 (3x) | Improved |
| use adjoint instead of direct sensitivity method | Algorithm 1, Pytorch · Cucker-Smale N=200 | time per epoch (s): 0.614 -> 2.819 | Hurt |
| replace Algorithm 1 with Algorithm 2 in Jax with JIT | Algorithm 1 adjoint, Jax · Cucker-Smale N=200 | time per epoch (s): 1.896 -> 0.0013 (x1458) | Improved |
| replace Algorithm 1 with Algorithm 3 in Jax | Algorithm 1 adjoint, Jax · Cucker-Smale N=5 | time per epoch (s): 0.935 -> 0.06 (x16) | Improved |
| replace Algorithm 1 with Algorithm 2 | Algorithm 1 direct, Pytorch · Cucker-Smale N=200 test | RSSE on test data: 4.32 × 10−7 -> 3.24 × 10−8 | Improved |
| replace Algorithm 1 with Algorithm 2 | Algorithm 1, Jax · Cucker-Smale N=200 test | RSSE on test data: 1.94 × 10−5 -> 1.46 × 10−7 | Improved |
| replace Algorithm 1 with Algorithm 3 | Algorithm 1, Jax · Cucker-Smale N=200 test | RSSE on test data: 1.94 × 10−5 -> 2.17 × 10−5 | No clear effect |
| replace Algorithm 1 with Algorithm 2 for 4950 gamma parameters | Algorithm 1 direct · Cucker-Smale N=200, non-uniform gamma | avg iteration time (s): 0.45 sec -> 0.06 sec | Improved |
| replace Algorithm 1 with Algorithm 2 for 4950 gamma parameters | Algorithm 1 direct · Cucker-Smale N=200, non-uniform gamma test | RSSE: 3.5 × 10−4 -> 3.6 × 10−4 | Hurt |
| replace Algorithm 1 with Algorithm 3 for black-box model | Algorithm 1 direct · Black-box neural ODE, Cucker-Smale N=200 | avg iteration time (s): 0.45 sec -> 0.29 sec | Improved |
Closest papers by meaning
- On Constraints in First-Order Optimization: A View from Non-Smooth Dynamical Systems (JMLR 2022 · similarity 0.88)
- Breaking the Convergence Barrier: Optimization via Fixed-Time Convergent Flows (AAAI 2022 · similarity 0.87)
- The Physical Systems Behind Optimization Algorithms (NeurIPS 2018 · similarity 0.87)
- TENG: Time-Evolving Natural Gradient for Solving PDEs With Deep Neural Nets Toward Machine Precision (ICML 2024 · similarity 0.86)
- Learning to Solve Differential Equation Constrained Optimization Problems (ICLR 2025 · similarity 0.85)
- Learning to Relax: Setting Solver Parameters Across a Sequence of Linear System Instances (ICLR 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.
