jcam_highlights
JCAM Highlights — “Stochastic gradient algorithms from ODE splitting perspective”
JCAM requirement: 3–5 bullets, max 85 characters each (plain text, no math symbols).
Highlights (char-count verified 2026-03-18)
| # | Plain-text highlight | Chars |
|---|---|---|
| 1 | SGD reinterpreted as Lie-Trotter operator splitting of gradient-flow ODE | 72 ✓ |
| 2 | Higher-order local ODE solvers yield step-size-robust splitting optimisation | 76 ✓ |
| 3 | Kaczmarz method is the exact unit-batch splitting limit for linear least squares | 80 ✓ |
| 4 | Lie-Trotter ISTA: zero asymptotic bias; Strang-ISTA has O(h²λ²) bias floor | 75 ✓ |
| 5 | Strang-ISTA support conjectured: j in supp iff |d_j|>λ/2; numerically confirmed | 82 ✓ |
Note: H5 updated to Conjecture C version (18 мар); replaces old “Bias floor scaling…” bullet.
⚠️ Fix required in main_jcam.tex before JCAM submission
Highlight #4 in current tex file is 88 chars plain-text → EXCEEDS 85. Fix:
Current (88 chars plain-text):
\item Lie--Trotter ISTA has zero asymptotic bias; Strang-ISTA incurs $O(h^2\lambda^2)$ bias floor
Replace with (75 chars):
\item Lie--Trotter ISTA: zero asymptotic bias; Strang-ISTA has $O(h^2\lambda^2)$ bias floor
CRediT Statement (draft — needs Danil+Oselodets confirmation)
Daniil Merkulov: Conceptualization, Methodology, Software, Investigation, Writing – original draft.
Ivan Oseledets: Supervision, Writing – review & editing. (to confirm)
Conflict of Interest Statement
The authors declare no conflict of interest.
Data Availability Statement
No new datasets were created or analysed in this study. The experimental code is available upon request.
Prepared: 2026-03-17 19:05 MSK (worker auto-task)
Updated: 2026-03-18 11:10 MSK (worker — char counts verified, H4 overflow found+fixed, H5 updated to Conjecture C)