TURBOTEST: Learning When Less is Enough through Early Termination of Internet Speed Tests
Researchers have developed TurboTest, a new framework designed to optimize internet speed tests by learning when to terminate them early without sacrificing accuracy. This approach decouples throughput prediction from the termination decision, using machine learning to estimate final speeds from partial measurements. Evaluated on over a million real-world speed tests, TurboTest demonstrated significant data savings compared to existing methods while maintaining accuracy. AI
IMPACT Optimizes internet speed tests using ML, potentially reducing data transfer costs for platforms and users.