diff options
| author | YurenHao0426 <Blackhao0426@gmail.com> | 2026-07-22 02:15:37 -0500 |
|---|---|---|
| committer | YurenHao0426 <Blackhao0426@gmail.com> | 2026-07-22 02:15:37 -0500 |
| commit | 1a22a99a5ebc857ae5a5270ffddacce6aa1386d8 (patch) | |
| tree | db61e1b9ea56f6512a6e66747f230b2bed7c119f /ROADMAP.md | |
| parent | 065c891be816b92a43065bafe89ad2aa266161b1 (diff) | |
experiments: predeclare low-query calibration screen
Diffstat (limited to 'ROADMAP.md')
| -rw-r--r-- | ROADMAP.md | 5 |
1 files changed, 5 insertions, 0 deletions
@@ -62,6 +62,11 @@ fixed epoch and fixed query/FLOP budgets. 16-directions-every-4-steps setting must preserve at least 90% of its gain over DFA. Pareto claims must hold in a hardware-independent cost coordinate as well as GTX-1080 wall time. +The validation screen is frozen to CIFAR-10 d20/w64, five epochs, seed 0, and compares the current +`K=16/every=4` reference with `K/every` pairs `1/4`, `2/8`, `4/16`, `1/8`, `1/16`, and `2/32`, plus +DFA. Selection uses final validation accuracy and logical batch loss queries; the test split is not +evaluated. The chosen low-query protocol then receives a multi-seed frozen confirmation. + ### C4. The comparison set contains the closest alternatives The main comparison must include exact BP, FA, DFA, direct/unamortized node perturbation, learned |