Fast Local Search
Algorithms AIMech:FastLocalSearch DRAFT
Aliases: FLS
A neighborhood-search procedure that uses activation bits to scan selected subneighborhoods, deactivating unsuccessful ones and reactivating those affected by accepted improvements.
Identity Notes & Concept Scope
FLS is the named Voudouris–Tsang neighborhood-reduction procedure, not any fast implementation of local search. It is separable from GLS feature-penalty updates and from the combined Guided Fast Local Search algorithm. Subneighborhoods can overlap; calling them a disjoint partition would overstate the source. No normalized identity match was found across all 442 pinned AIO labels/scoped synonyms and 304 preceding native IDs/names/aliases. Full-checkout fixed-string searches included ignored and hidden files in worktree and main, returning 176 and 176 lines respectively. Inspected text hits were prior discovery leads, uses of local search inside existing methods, source references and their saved reports/reviews. These were references to methods rather than existing candidate records. Six exact AIO CLASS nodes with scoped synonyms/comments and outgoing edges and six native neighboring mechanisms were inspected. Absence is bounded to this snapshot and semantic neighborhood.
Claims & Grounded Evidence
1 Claim
Mechanism
#method
Decompose the neighborhood into indexed, possibly overlapping subneighborhoods and supply an activation bit for each. The standalone initialization activates all; the callable routine also permits a supplied subset. Repeatedly scan in the declared order, skipping inactive subneighborhoods. In an active subneighborhood, take the first encountered move that strictly improves the current objective, update the solution, retain that subneighborhood as active, and activate the configured subneighborhoods affected by the move. Continue with the next subneighborhood instead of restarting the entire scan. If no improving move is found after checking the current subneighborhood, deactivate it. Return the current solution when all bits are inactive. In the optional GLS combination, changes to feature penalties reactivate associated subneighborhoods before the next FLS call; the penalty formula itself belongs to GLS.
⚠️ Curated Limitations & Negative Results
- All bits inactive certifies only the termination of the configured activation scheme. If initial activation or later reactivation omits a region where an improvement exists or becomes possible, the result need not be a local optimum of the full neighborhood. The authors explicitly call the result approximate and warn that improvements can be missed.
- Activation dependencies, move coverage, order and overlap are problem-dependent. Reactivating every subneighborhood is conservative but can remove the intended computational savings; the method is not a universal speedup guarantee.
- Within one call the objective is fixed and accepted moves strictly improve it. Curator deduction: finite states and finite move scans imply termination, but neither arbitrary continuous termination nor overall GLS termination follows. Penalty changes between GLS calls alter the objective and require renewed activation.
Artificial Intelligence Ontology (AIO) Grounding
Explicit AIO Snapshot Assessment: No Match
Rationale: No suitable AIO target was found for this general neighborhood-search method within the inspected snapshot and semantic neighborhood. Optional model-training applications or objective evaluation do not establish a learning-task or model identity.
Search Scope: All 442 pinned AIO identities/scoped synonyms and 304 prior native IDs/names/aliases; ignored/hidden text in both checkouts; semantic label search for search, gradient and optimization; six exact AIO nodes and six native neighbors.
| Examined AIO Candidate | Rejection Reason |
|---|---|
AIO:TrainingStrategy
|
The definition concerns preprocessing or training machine-learning models. General local-search optimization is not restricted to such training. |
AIO:MachineLearningTask
|
The target concerns learning methods from data; minimizing a supplied combinatorial objective is not necessarily that task. |
AIO:ReinforcementLearning
|
Memory, adaptive penalties or an acceptance criterion do not themselves define a policy/environment learning problem. Exploration terminology is insufficient for a forced mapping. |
AIO:ReinforcementLearningLLM
|
The method is not a language model fine-tuned through reward feedback. |
AIO:EvolutionaryLLM
|
The target is a large language model. Its related evolutionary-algorithms synonym does not identify all optimizers or convert a single-trajectory method into a model. |
AIO:MarkovChain
|
Some memoryless configurations admit a Markov description, while others use search history. That mathematical representation alone does not identify the optimization framework with the pinned network/model class or its semantic parent. |
Evidence Sources & Primary Literature
2 Sources| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
method-source
|
Christos Voudouris, Edward P. K. Tsang, Abdullah Alsheddy
|
primary paper
full_text
|
ID: doi:10.1007/978-1-4419-1665-5_11
sha256:e4dd9aed0fcf…
|
additional-1
|
Christos Voudouris, Edward Tsang
|
primary paper
full_text
|
ID: doi:10.1016/S0377-2217(98)00099-X
sha256:4550ebdad45d…
|
Curation Provenance
Curator: Codex
Method: ai_assisted
Model: GPT-6
Created: 2026-10-11T14:27:03Z
Updated: 2026-10-11T14:27:03Z
Source File:
data/curated/FastLocalSearch.yaml