ALEXANDRIA, Va., Aug. 12 -- United States Patent no. 12,705,547, issued on Aug. 11.
"Dual-time-scale adaptive architecture for reliable online learning" was invented by R. David Brown (Palm Springs, Calif.).
According to the abstract* released by the U.S. Patent & Trademark Office: "A computing system operates a machine learning model comprising base parameters and adapter parameters applied as additive low-rank deltas to one or more transformations. The system determines a routing certificate for each request using a scope function and applies only an adapter state keyed to the determined routing certificate during inference. When a supervision signal is received, the system computes a candidate update and evaluates a gating policy that ...