Adadelta
Algorithms AIMech:Adadelta DRAFT
An adaptive gradient optimization method that rescales gradients using exponentially weighted squared-gradient and squared-update histories.
Identity Notes & Concept Scope
The additional update-history scale distinguishes Adadelta from RMSprop and from Adagrad's cumulative accumulator.
Claims & Grounded Evidence
1 Claim
Mechanism
#update-rule
The update multiplies the gradient by the stabilized root-mean-square of prior updates divided by the stabilized root-mean-square of current gradient history, then updates the squared-update average.
⚠️ Curated Limitations & Negative Results
- The inspected implementation includes a learning-rate multiplier; this record does not describe it as parameter-free.
- No convergence theorem or comparative benchmark result is curated.
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
|
sha256:6741996fc490…
|
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/Adadelta.yaml