AWhat changed
Uses a reinforcement-learned controller to sample subgraphs of a shared computation graph, alternating controller validation rewards with shared-weight training to reduce repeated child-model training.
Scientific scope
Uses a reinforcement-learned controller to sample subgraphs of a shared computation graph, alternating controller validation rewards with shared-weight training to reduce repeated child-model training. A learned controller selected network subgraphs using validation rewards; human authors defined the shared model, training procedure and tests. Shared-weight validation is a proxy for independently retrained performance; selected architectures are retrained from scratch. Reported search-cost factors do not mean zero search/final-training cost or globally optimal architectures. Pham and Guan are equal contributors.
Evidence
A learned controller selected network subgraphs using validation rewards; human authors defined the shared model, training procedure and tests. WRS has not independently rerun these experiments. A learned controller selected network subgraphs using validation rewards; human authors defined the shared model, training procedure and tests. Shared-weight validation is a proxy for independently retrained performance; selected architectures are retrained from scratch. Reported search-cost factors do not mean zero search/final-training cost or globally optimal architectures. Pham and Guan are equal contributors.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction