AdaMax
Algorithms AIMech:AdaMax DRAFT
An Adam-family optimizer using an exponentially weighted infinity-norm gradient scale instead of a second raw moment denominator.
Identity Notes & Concept Scope
AdaMax changes Adam's normalization statistic and is retained as a distinct algorithm; it is not an alias for Adam.
Claims & Grounded Evidence
1 Claim
Mechanism
#update-rule
The first-moment estimate is bias-corrected, while the denominator takes the coordinatewise maximum of the decayed prior scale and the current absolute gradient with the implementation's stabilizer.
⚠️ Curated Limitations & Negative Results
- Epsilon placement follows the cited PyTorch implementation.
- No universal convergence or benchmark advantage is asserted.
Artificial Intelligence Ontology (AIO) Grounding
Mapped AIO Classes
| AIO Target Class | Mapping Relation | Mapping Rationale |
|---|---|---|
AIO:TrainingStrategy
TrainingStrategy
|
related_match | This algorithm can implement parameter fitting, but the pinned TrainingStrategy definition uses a preprocessing genus. This is a contextual training relationship, not equivalence or a subclass assertion. |
Evidence Sources & Primary Literature
1 Source| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
pytorch-docs
|
PyTorch contributors
|
official documentation
documentation
|
sha256:4a069f9b4d56…
|
Curation Provenance
Curator: Codex
Method: ai_assisted
Model: GPT-6
Created: 2026-10-11T03:25:18Z
Updated: 2026-10-11T03:25:18Z
Source File:
data/curated/AdaMax.yaml