Random Online Aggressive Racing
Algorithms AIMech:RandomOnlineAggressiveRacing DRAFT
Aliases: ROAR, Random Online Adaptive Racing
An algorithm-configuration procedure that proposes uniformly random parameter settings and compares each challenger with an incumbent through adaptive evaluations on shared problem instances and random seeds, without fitting a predictive performance model.
Identity Notes & Concept Scope
The 2011 source expands ROAR as Random Online Adaptive Racing in its introduction and Random Online Aggressive Racing in Section3. Both wordings here refer to that source’s same method. It shares SMAC’s intensification procedure but omits SMAC’s random-forest surrogate and acquisition-guided proposals. It differs from fixed-evaluation random search and significance-based elimination in F-Race. No normalized identity match was found across all 442 pinned AIO labels/scoped synonyms and 265 preceding native IDs/names/aliases. Full-checkout fixed-string searches included ignored and hidden files in both worktree and main checkout, returning 20 lines in the worktree and 25 in main. These were prior candidate leads, source references, neighboring-method mentions and five unrelated ROAR namespace entries in ignored dependencies, not existing records for these identities. Seven exact AIO CLASS nodes with synonyms and outgoing edges and six neighboring native mechanisms were inspected. Absence is bounded to this snapshot and semantic neighborhood.
Claims & Grounded Evidence
1 Claim
Mechanism
#method
Initialize by evaluating the default configuration or a random setting when a default is unavailable. Supply uniformly random configurations to the intensification routine; the nominal model is constant and unused for proposing them. Add an incumbent evaluation, subject to the maximum-run cap, on an instance with the fewest incumbent runs and a random seed. Evaluate a challenger on a subset of the incumbent’s instance/seed pairs and compare their empirical costs on exactly those pairs. Reject a worse challenger; otherwise enlarge its tested subset in doubling batches until it has covered the incumbent’s available runs, then replace the incumbent when the challenger remains at least as good. Continue proposing random challengers within the configuration budget.
⚠️ Curated Limitations & Negative Results
- Aggressive racing in this source makes empirical decisions without waiting for a statistical significance threshold. An early rejection can discard a configuration with better expected performance; no family-wise error control or finite-budget optimum is claimed.
- The shared intensification routine caps incumbent runs at maxR=2000. Its time-budget early-exit guard requires i>=2, but ROAR supplies one challenger per call, so the proposal loop ends after that one challenger; the guard is not a mandatory two-challenger minimum for ROAR. The paper’s asymptotic discussion assumes consistent estimation and unbounded evaluation growth in a finite configuration space, which are not automatic properties of the capped practical routine.
- The original procedure lacks adaptive capping of individual poor runs. Uniform configuration proposals do not mean equal allocation of evaluations to every configuration. No implementation or benchmark outcome was reproduced.
Artificial Intelligence Ontology (AIO) Grounding
Explicit AIO Snapshot Assessment: No Match
Rationale: No inspected AIO target captures this general algorithm-configuration procedure or a defining component. The local algorithm remains supported by its primary source; this is a bounded mapping assessment, not absence from all ontologies.
Search Scope: All 442 pinned AIO CLASS identities and scoped synonyms, all 265 preceding maintained identities, ignored/hidden checkout text, and semantic inspection of seven AIO targets plus neighboring configuration/resource-allocation methods.
| Examined AIO Candidate | Rejection Reason |
|---|---|
AIO:TrainingStrategy
|
The pinned definition concerns preprocessing/training of machine-learning models. This method configures target algorithms generally; an optional use for model tuning does not establish a training-strategy identity or necessary relation. |
AIO:MachineLearningTask
|
The target is a broad process for building methods that learn from data. Configuring an arbitrary solver over instances need not build such a learning method; shared use of observations does not justify equivalence or subclassing. |
AIO:MetaLearning
|
The pinned task learns from metadata about machine-learning experiments. These methods require no collection of ML experiments or learned cross-task prior; solver configuration alone is not that identity. |
AIO:ActiveLearning
|
The pinned task queries labels for data points. Choosing configuration evaluations or instance runs is not label acquisition in that defined sense. |
AIO:ReinforcementLearning
|
Exploration and empirical feedback do not by themselves establish the pinned learning-task identity; no reward-learning agent or policy-learning component is required by these configurators. |
AIO:RandomForest
|
These methods contain no defining decision-tree ensemble surrogate; random sampling, randomized local search and rank tests are not random forests. |
AIO:EvaluationBias
|
This is a bias concept, not a configuration algorithm. Potential overfitting or distribution mismatch is a limitation rather than an identity mapping. |
Evidence Sources & Primary Literature
1 Source| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
method-source
|
Frank Hutter, Holger H. Hoos, Kevin Leyton-Brown
|
primary paper
full_text
|
ID: doi:10.1007/978-3-642-25566-3_40
sha256:aecf80aba313…
|
Curation Provenance
Curator: Codex
Method: ai_assisted
Model: GPT-6
Created: 2026-10-11T12:04:49Z
Updated: 2026-10-11T12:09:15Z
Source File:
data/curated/RandomOnlineAggressiveRacing.yaml