inactive
service state
RESEARCH COMPUTE
ML Lab
Rule, ridge, and LightGBM challengers under chronological, cost-aware, fail-closed validation.
INACTIVE
01
Training runtime
inactive · success · next Fri 2026-09-11 03:33:41 PDT
success
last result
Fri 2026-09-11 03:33:41 PDT
next scheduled run
2 cores
bounded ML CPU ceiling
02
Latest retained run
never · 0 retained evaluations across 0 completed runs
never
run identity
0
completed training runs
0
retained evaluations
0
active research lanes
| Pair | Model | Fold | Events | Net mean | MAE | Disposition |
|---|---|---|---|---|---|---|
| Detailed model evidence remains in the retained research record; this public view publishes bounded lifecycle totals. | ||||||
03
Validation boundaries
Failures remain visible and cannot silently promote a model.
Chronology first
Prefix and recursive-startup comparisons must pass their frozen tolerances before fitting.
Costs included
Evaluations retain the declared round-trip cost and abstain where inputs are ineligible.
No automatic promotion
Training artifacts are exploratory evidence. They do not authorize live trading or establish future profit.
A running job proves compute activity only. A completed report still requires artifact, temporal, drift, calibration, and human review.