Research map / Optimization / Nonconvex and smooth optimization
Optimal Stochastic and Online Learning with Individual Iterates
Yunwen Lei, Peng Yang, Ke Tang, Ding-Xuan Zhou
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 OCMDI individual iterate instead of uniform averaging (ablation) | SCMD output · Binary classification test sets (SGD, SPGD) | test objective value | Improved |
| Use suffix averaging instead of weighted averaging (ablation) | SCMD output · Binary classification test sets | test objective value | Improved |
| Use OCMDI instead of random-index output (RAND) (ablation) | SCMD output · Binary classification test sets | test objective value | Improved |
| Use OCMDI instead of last iterate (LAST) (ablation) | SCMD output · Binary classification test sets | test objective value | No clear effect |
| Use OCMDI instead of uniform averaging (ablation) | SCMD output · Tomography reconstruction N=64 | number of non-zero components: 4096 -> 1828 | Improved |
| Use OCMDI instead of suffix averaging (ablation) | SCMD output · Tomography reconstruction N=64 | number of non-zero components: 1943 -> 1828 | Improved |
| Use OCMDI instead of weighted averaging (ablation) | SCMD output · Tomography reconstruction N=64 | number of non-zero components: 3727 -> 1828 | Improved |
| Use suffix averaging output instead of random-index output (RAND) (ablation) | SCMD output · Tomography reconstruction | training speed | Improved |
| Choose T* as argmin of A_t over [T,2T-1] instead of Algorithm 1 rule (ablation) | OCMDI · Experiments | performance | Hurt |
Closest papers by meaning
- Stochastic Composite Mirror Descent: Optimal Bounds with High Probabilities (NeurIPS 2018 · similarity 0.91)
- Sparse Learning with Stochastic Composite Optimization (TPAMI 2016 · similarity 0.90)
- Efficient Mirror Descent Ascent Methods for Nonsmooth Minimax Problems (NeurIPS 2021 · similarity 0.89)
- Stochastic Gradient/Mirror Descent: Minimax Optimality and Implicit Regularization (ICLR 2019 · similarity 0.88)
- Online mirror descent and dual averaging: keeping pace in the dynamic case (ICML 2020 · similarity 0.88)
- The Statistical Complexity of Early-Stopped Mirror Descent (NeurIPS 2020 · similarity 0.88)
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.
