Policy-Controlled Generalized Share: A General Framework with a Transformer Instantiation for Strictly Online Switching-Oracle Tracking

Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• Policy-Controlled Generalized Share (PCGS) is a general strictly online framework where the generalized-share recursion is fixed but post-loss update controls can vary adaptively.method 95%
• PCGS-TF, which uses a causal Transformer as an update controller, achieves the lowest mean dynamic regret in all seven non-stationary families on a controlled synthetic suite.result 90%
• PCGS-TF achieves the lowest normalized dynamic regret for S = 5, 10, and 20 on a reproduced household-electricity benchmark.result 90%
• Under admissible post-loss update controls, PCGS obtains a pathwise weighted regret guarantee for general time-varying learning rates.technical 85%
• PCGS obtains a standard dynamic-regret guarantee against any expert path with at most S switches under the constant-learning-rate specialization.technical 85%
• The PCGS framework separates the analysis of the share backbone from the design of the controller.method 80%
• The update in PCGS exposes three distinct control channels: learning-rate control (ηt), restart intensity (ρt), and restart distribution (qt).technical 95%
• PCGS can be applied in high-stakes settings such as finance, operations, and safety monitoring, which requires responsible use practices.market 75%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card